Curve Fitting

Toolbox
For Use with MATLAB ®

Computation

Visualization

Programming

User’s Guide
Version 1
How to Contact The MathWorks:

www.mathworks.com Web
comp.soft-sys.matlab Newsgroup

[email protected] Technical support

[email protected] Product enhancement suggestions
[email protected] Bug reports
[email protected] Documentation error reports
[email protected] Order status, license renewals, passcodes
[email protected] Sales, pricing, and general information

508-647-7000 Phone

508-647-7001 Fax

The MathWorks, Inc. Mail

3 Apple Hill Drive
Natick, MA 01760-2098

For contact information about worldwide offices, see the MathWorks Web site.

Curve Fitting Toolbox User’s Guide

 COPYRIGHT 2001-2002 by The MathWorks, Inc.
The software described in this document is furnished under a license agreement. The software may be used
or copied only under the terms of the license agreement. No part of this manual may be photocopied or repro-
duced in any form without prior written consent from The MathWorks, Inc.
FEDERAL ACQUISITION: This provision applies to all acquisitions of the Program and Documentation by
or for the federal government of the United States. By accepting delivery of the Program, the government
hereby agrees that this software qualifies as "commercial" computer software within the meaning of FAR
Part 12.212, DFARS Part 227.7202-1, DFARS Part 227.7202-3, DFARS Part 252.227-7013, and DFARS Part
252.227-7014. The terms and conditions of The MathWorks, Inc. Software License Agreement shall pertain
to the government’s use and disclosure of the Program and Documentation, and shall supersede any
conflicting contractual terms or conditions. If this license fails to meet the government’s minimum needs or
is inconsistent in any respect with federal procurement law, the government agrees to return the Program
and Documentation, unused, to MathWorks.
MATLAB, Simulink, Stateflow, Handle Graphics, and Real-Time Workshop are registered trademarks, and
TargetBox is a trademark of The MathWorks, Inc.
Other product or brand names are trademarks or registered trademarks of their respective holders.
Printing History: July 2001 First printing New for Version 1 (Release 12.1)
July 2002 Second printing Revised for Version 1.1 (Release 13)
 Contents

Preface

What Is the Curve Fitting Toolbox? . . . . . . . . . . . . . . . . . . . . . . vi

Exploring the Toolbox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi

Related Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

Using This Guide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

Expected Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
How This Guide Is Organized . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Documentation Examples and Data Sets . . . . . . . . . . . . . . . . . . . . x

Installation Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

Typographical Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii

Getting Started with the Curve Fitting Toolbox

1
Opening the Curve Fitting Tool . . . . . . . . . . . . . . . . . . . . . . . . 1-2

Importing the Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-3

Fitting the Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-5

The Data Fitting Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-5
Determining the Best Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-8
Saving the Fit Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-13

Analyzing the Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-15

Saving the Analysis Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-16

i
 Saving Your Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-17
Saving the Session . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-18
Generating an M-File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-19

Importing, Viewing, and Preprocessing Data

2
Importing Data Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2
Example: Importing Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-4

Viewing Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-6

Viewing Data Graphically . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-6
Viewing Data Numerically . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-8

Smoothing Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-9

Moving Average Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-12
Lowess and Loess: Local Regression Smoothing . . . . . . . . . . . 2-14
Savitzky-Golay Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-19
Example: Smoothing Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-21

Excluding and Sectioning Data . . . . . . . . . . . . . . . . . . . . . . . . 2-25

Marking Outliers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-27
Sectioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-30
Example: Excluding and Sectioning Data . . . . . . . . . . . . . . . . 2-32
Example: Sectioning Periodic Data . . . . . . . . . . . . . . . . . . . . . . 2-35

Additional Preprocessing Steps . . . . . . . . . . . . . . . . . . . . . . . 2-40

Transforming the Response Data . . . . . . . . . . . . . . . . . . . . . . . 2-40
Removing Infs, NaNs, and Outliers . . . . . . . . . . . . . . . . . . . . . 2-41

Selected Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-42

ii Contents
 Fitting Data
3
The Fitting Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-2

Parametric Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-4

Basic Assumptions About the Error . . . . . . . . . . . . . . . . . . . . . . 3-5
The Least Squares Fitting Method . . . . . . . . . . . . . . . . . . . . . . . 3-6
Library Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-16
Custom Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-20
Specifying Fit Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-23
Evaluating the Goodness of Fit . . . . . . . . . . . . . . . . . . . . . . . . . 3-27
Example: Rational Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-41
Example: Fitting with Custom Equations . . . . . . . . . . . . . . . . 3-46
Example: Robust Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-61

Nonparametric Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-68

Interpolants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-68
Smoothing Spline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-70
Example: Nonparametric Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-72

Selected Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-75

Function Reference
4
Functions – By Category . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-78
Fitting Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-78
Getting Information and Help . . . . . . . . . . . . . . . . . . . . . . . . . . 4-78
Getting and Setting Properties . . . . . . . . . . . . . . . . . . . . . . . . . 4-78
Preprocessing Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-78
Postprocessing Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-79
General Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-79

Functions – Alphabetical List . . . . . . . . . . . . . . . . . . . . . . . . . 4-80

iii
iv Contents
Preface

This chapter provides an overview of the Curve Fitting Toolbox, as well as information about this
documentation. The sections are as follows.

What Is the Curve Fitting The toolbox and the kinds of tasks it can perform
Toolbox? (p. vi)
Related Products (p. viii) MathWorks products related to this toolbox
Using This Guide (p. ix) An overview of this guide
Installation Information How to determine whether the toolbox is installed on your system
(p. xi)
Typographical Conventions Typographical conventions that this guide uses
(p. xii)
Preface

What Is the Curve Fitting Toolbox?

The Curve Fitting Toolbox is a collection of graphical user interfaces (GUIs)
and M-file functions built on the MATLAB® technical computing environment.
The toolbox provides you with these main features:

• Data preprocessing such as sectioning and smoothing

• Parametric and nonparametric data fitting:
- You can perform a parametric fit using a toolbox library equation or using
a custom equation. Library equations include polynomials, exponentials,
rationals, sums of Gaussians, and so on. Custom equations are equations
that you define to suit your specific curve fitting needs.
- You can perform a nonparametric fit using a smoothing spline or various
interpolants.
• Standard linear least squares, nonlinear least squares, weighted least
squares, constrained least squares, and robust fitting procedures
• Fit statistics to assist you in determining the goodness of fit
• Analysis capabilities such as extrapolation, differentiation, and integration
• A graphical environment that allows you to:
- Explore and analyze data sets and fits visually and numerically
- Save your work in various formats including M-files, binary files, and
workspace variables

Exploring the Toolbox

The Curve Fitting Toolbox consists of two different environments: a graphical
user interface (GUI) environment and the usual MATLAB command line
environment.
Although the two environments are functionally equivalent, you generally
cannot mix the two when performing a given curve fitting task. For example,
you cannot generate a fit at the command line and then import that fit into the
graphical environment. However, you can create a fit in the graphical
environment and then generate an associated M-file. You can then recreate the
fit from the command line and modify the M-file according to your needs. For
this reason, as well as for the enhanced data analysis and exploration tools that
are available, we recommend you use the graphical environment for most
tasks.

vi
 What Is the Curve Fitting Toolbox?

You can explore the graphical environment by typing

cftool

Click the GUI Help buttons to learn how to proceed. Additionally, you can
follow the examples in the tutorial sections of this guide, which are all GUI
oriented.
To explore the command line environment, you can list the toolbox functions by
typing
help curvefit

To view the code for any function, type

type function_name

To view the help for any function, type

help function_name

You can change the way any toolbox function works by copying and renaming
the M-file, and then modifying your copy. However, these changes will not be
reflected in the graphical environment.
You can also extend the toolbox by adding your own M-files, or by using it in
combination with other products such as the Statistics Toolbox or the
Optimization Toolbox.

vii
Preface

Related Products
The Curve Fitting Toolbox requires MATLAB 6.5 (Release 13). Additionally,
The MathWorks provides several related products that are especially relevant
to the kinds of tasks you can perform with the Curve Fitting Toolbox. For more
information about any of these products, see either

• The online documentation for that product if it is installed or if you are

reading the documentation from the CD
• The MathWorks Web site, at http://www.mathworks.com; see the “products”
section

Note The toolboxes listed below all include functions that extend the
capabilities of MATLAB.

Product Description

Optimization Toolbox Solve standard and large-scale optimization

problems

Spline Toolbox Create and manipulate spline approximation

models of data

Statistics Toolbox Apply statistical algorithms and probability

models

viii
 Using This Guide

Using This Guide

Expected Background
This guide assumes that you already have background knowledge in the
subject of curve fitting data. If you do not yet have this background, then you
should read a standard curve fitting text, some of which are listed in “Selected
Bibliography” on page 3-75.
When learning to use the toolbox, you should start with Chapter 1, “Getting
Started with the Curve Fitting Toolbox,” which illustrates the major toolbox
features by fitting census data to several equations and then evaluating the
best fit. More detailed information about the features discussed in this chapter
is available elsewhere in the guide, and you should browse the documentation
according to your needs.

How This Guide Is Organized

The organization of this guide is described below.

Chapter Description

Getting Started with the Describes a particular example in detail to help

Curve Fitting Toolbox you get started with the toolbox.

Importing, Viewing, and Describes how to import, view, and smooth

Preprocessing Data data sets, and how to exclude one or more data
points in a data set from subsequent fits.

Fitting Data Describes how to fit data using library or

custom equations, a smoothing spline, or an
interpolant. Several detailed examples are
included as well.

Function Reference Describes all toolbox functions in detail.

ix
Preface

Documentation Examples and Data Sets

To learn how to use the Curve Fitting Toolbox, you can follow the examples
included in this guide. A quick way to locate these examples is with the
example index, which you can access via the Help browser.
Some examples use data that is generated as part of the example, while other
examples use data sets that are included with the toolbox or with MATLAB.
These data sets are stored as MAT-files and are listed below.

Data Set Description

carbon12alpha Generated data associated with the nuclear reaction

12
C(e,e’,α)8Be. The data set is based on the experiment
described by reference [8] in Chapter 3, “Fitting Data.”

census United States population figures for the years 1790 to

1990 in 10-year increments. The data set is included
with MATLAB.

enso Measured data of monthly averaged atmospheric

pressure differences between Easter Island and Darwin,
Australia.

flvote2k 2000 U. S. presidential election results for the state of

Florida. The data set was obtained from
http://election.dos.state.fl.us/.

gauss3 Generated data consisting of two poorly resolved

Gaussian peaks on an exponential background.

hahn1 Measured data of the thermal expansion of copper as a

function of temperature.

The enso, gauss3, and hahn1 data sets were obtained from the Statistical
Reference Datasets collection, which is provided by the Information
Technology Laboratory of the National Institute of Standards and Technology.
The Web site is http://www.itl.nist.gov/div898/strd.

x
 Installation Information

Installation Information
To determine if the Curve Fitting Toolbox is installed on your system, type
ver

at the MATLAB prompt. MATLAB displays information about the version of

MATLAB you are running, including a list of installed add-on products and
their version numbers. Check the list to see if the Curve Fitting Toolbox
appears.
For information about installing the toolbox, refer to the MATLAB Installation
Guide for your platform. If you experience installation difficulties and have
Web access, look for the installation and license information at the MathWorks
Web site http://www.mathworks.com/support.

xi
Preface

Typographical Conventions
This guide uses some or all of these conventions.

Item Convention Example

Example code Monospace font To assign the value 5 to A,

enter
A = 5

Function names, syntax, Monospace font The cos function finds the
filenames, directory/folder cosine of each array element.
names, and user input Syntax line example is
MLGetVar ML_var_name

Buttons and keys Boldface with book title caps Press the Enter key.

Literal strings (in syntax Monospace bold for literals f = freqspace(n,'whole')

descriptions in reference
chapters)

Mathematical Italics for variables This vector represents the

expressions Standard text font for functions, polynomial p = x2 + 2x + 3.
operators, and constants

MATLAB output Monospace font MATLAB responds with

A =
5

Menu and dialog box titles Boldface with book title caps Choose the File Options
menu.

New terms and for Italics An array is an ordered

emphasis collection of information.

Omitted input arguments (...) ellipsis denotes all of the [c,ia,ib] = union(...)
input/output arguments from
preceding syntaxes.

String variables (from a Monospace italics sysc = d2c(sysd,'method')

finite list)

xii
 1
Getting Started with the
Curve Fitting Toolbox

This chapter describes a particular example in detail to help you get started with the Curve Fitting
Toolbox. In this example, you will fit census data to several toolbox library models, find the best fit,
and extrapolate the best fit to predict the US population in future years. In doing so, the basic steps
involved in any curve fitting scenario are illustrated. These steps include

Opening the Curve Fitting The Curve Fitting Tool is the main toolbox interface.
Tool (p. 1-2)
Importing the Data (p. 1-3) The data must exist as vectors in the MATLAB workspace. After
importing, you can view the data, mark data points to be excluded
from the fit, and smooth the data.
Fitting the Data (p. 1-5) Explore various parametric and nonparametric fits, and compare fit
results graphically and numerically.
Analyzing the Fit (p. 1-15) Evaluate (interpolate or extrapolate), differentiate, or integrate the fit.
Saving Your Work (p. 1-17) Save your work for documentation purposes or for later analysis.
1 Getting Started with the Curve Fitting Toolbox

Opening the Curve Fitting Tool

The Curve Fitting Tool is a graphical user interface (GUI) that allows you to

• Visually explore one or more data sets and fits as scatter plots.
• Graphically evaluate the goodness of fit using residuals and prediction
bounds.
• Access additional interfaces for
- Importing, viewing, and smoothing data
- Fitting data, and comparing fits and data sets
- Marking data points to be excluded from a fit
- Selecting which fits and data sets are displayed in the tool
- Interpolating, extrapolating, differentiating, or integrating fits
You open the Curve Fitting Tool with the cftool command.
cftool

1-2
 Importing the Data

Importing the Data

Before you can import data into the Curve Fitting Tool, the data variables must
exist in the MATLAB workspace. For this example, the data is stored in the file
census.mat, which is provided with MATLAB.
load census

The workspace now contains two new variables, cdate and pop:

• cdate is a column vector containing the years 1790 to 1990 in 10-year

increments.
• pop is a column vector with the US population figures that correspond to the
years in cdate.

You can import data into the Curve Fitting Tool with the Data GUI. You open
this GUI by clicking the Data button on the Curve Fitting Tool. As shown
below, the Data GUI consists of two panes: Data sets and Smooth. The Data
Sets pane allows you to

• Import predictor (X) data, response (Y) data, and weights. If you do not
import weights, then they are assumed to be 1 for all data points.
• Specify the name of the data set.
• Preview the data.

1-3
1 Getting Started with the Curve Fitting Toolbox

To load cdate and pop into the Curve Fitting Tool, select the appropriate
variable names from the X Data and Y Data lists. The data is then displayed
in the Preview window. Click the Create data set button to complete the data
import process.

Select the data

variable names.

Click Create data set

to import the data.

The Smooth pane is described in Chapter 2, “Importing, Viewing, and

Preprocessing Data.”

1-4
 Fitting the Data

Fitting the Data

You fit data with the Fitting GUI. You open this GUI by clicking the Fitting
button on the Curve Fitting Tool. The Fitting GUI consists of two parts: the Fit
Editor and the Table of Fits. The Fit Editor allows you to

• Specify the fit name, the current data set, and the exclusion rule.
• Explore various fits to the current data set using a library or custom
equation, a smoothing spline, or an interpolant.
• Override the default fit options such as the coefficient starting values.
• Compare fit results including the fitted coefficients and goodness of fit
statistics.

The Table of Fits allows you to

• Keep track of all the fits and their data sets for the current session.
• Display a summary of the fit results.
• Save or delete the fit results.

The Data Fitting Procedure

For this example, begin by fitting the census data with a second degree
polynomial. Then continue fitting the data using polynomial equations up to
sixth degree, and a single-term exponential equation.
The data fitting procedure follows these general steps:

1 From the Fit Editor, click New Fit.

Note that this action always defaults to a linear polynomial fit type. You use
New Fit at the beginning of your curve fitting session, and when you are
exploring different fit types for a given data set.

2 Because the initial fit uses a second degree polynomial, select quadratic
polynomial from the Polynomial list. Name the fit poly2.

3 Click the Apply button or select the Immediate apply check box. The
library model, fitted coefficients, and goodness of fit statistics are displayed
in the Results area.

1-5
1 Getting Started with the Curve Fitting Toolbox

4 Fit the additional library equations.

For fits of a given type (for example, polynomials), you should use Copy Fit
instead of New Fit because copying a fit retains the current fit type state
thereby requiring fewer steps than creating a new fit each time.
The Fitting GUI is shown below with the results of fitting the census data with
a quadratic polynomial.

The Fit Editor allows you to select a data

set and a fit name, and to explore and
compare various library and custom fits.

The Table of Fits allows you to keep

track of all the fits, their data sets,
and fit results for the current session.

1-6
 Fitting the Data

The data, fit, and residuals are shown below. You display the residuals as a line
plot by selecting the menu item View->Residuals->Line plot from the Curve
Fitting Tool.

These residuals indicate that

a better fit may be possible.

The residuals indicate that a better fit may be possible. Therefore, you should
continue fitting the census data following the procedure outlined in the
beginning of this section.
The residuals from a good fit should look random with no apparent pattern. A
pattern, such as a tendency for consecutive residuals to have the same sign, can
be an indication that a better model exists.

1-7
1 Getting Started with the Curve Fitting Toolbox

When you fit higher degree polynomials, the Results area displays this
warning:
Equation is badly conditioned. Remove repeated data points
or try centering and scaling.

The warning arises because the fitting procedure uses the cdate values as the
basis for a matrix with very large values. The spread of the cdate values
results in scaling problems. To address this problem, you can normalize the
cdate data. Normalization is a process of scaling the predictor data to improve
the accuracy of the subsequent numeric computations. A way to normalize
cdate is to center it at zero mean and scale it to unit standard deviation.
(cdate - mean(cdate))./std(cdate)

To normalize data with the Curve Fitting Tool, select the Center and scale X
data check box.

Note Because the predictor data changes after normalizing, the values of the
fitted coefficients also change when compared to the original data. However,
the functional form of the data and the resulting goodness of fit statistics do
not change. Additionally, the data is displayed in the Curve Fitting Tool using
the original scale.

Determining the Best Fit

To determine the best fit, you should examine both the graphical and
numerical fit results.

Examining the Graphical Fit Results

Your initial approach in determining the best fit should be a graphical
examination of the fits and residuals. The graphical fit results shown below
indicate that

• The fits and residuals for the polynomial equations are all similar, making it
difficult to choose the best one.

1-8
 Fitting the Data

• The fit and residuals for the single-term exponential equation indicate it is a
poor fit overall. Therefore, it is a poor choice for extrapolation.

To easily view all the data,

fits, and residuals, turn
the legend off.

The residuals for the polynomial

fits are all similar making it
difficult to choose the best one.

The residuals for the

exponential fit indicate
it is a poor fit overall.

Use the Plotting GUI to remove exp1 from the scatter plot display.

Remove this fit from

the scatter plot.

1-9
1 Getting Started with the Curve Fitting Toolbox

Because the goal of fitting the census data is to extrapolate the best fit to
predict future population values, you should explore the behavior of the fits up
to the year 2050. You can change the axes limits of the Curve Fitting Tool by
selecting the menu item Tools->Axes Limit Control.
The census data and fits are shown below for an upper abscissa limit of 2050.
The behavior of the sixth degree polynomial fit beyond the data range makes it
a poor choice for extrapolation.

The sixth degree polynomial fit

beyond the data range makes it
a poor choice for extrapolation.

Change the upper abscissa

limit to 2050.

As you can see, you should exercise caution when extrapolating with
polynomial fits because they can diverge wildly outside the data range.

1-10
 Fitting the Data

Examining the Numerical Fit Results

Because you can no longer eliminate fits by examining them graphically, you
should examine the numerical fit results. There are two types of numerical fit
results displayed in the Fitting GUI: goodness of fit statistics and confidence
intervals on the fitted coefficients. The goodness of fit statistics help you
determine how well the curve fits the data. The confidence intervals on the
coefficients determine their accuracy.
Some goodness of fit statistics are displayed in the Results area of the Fit
Editor for a single fit. All goodness of fit statistics are displayed in the Table
of Fits for all fits, which allows for easy comparison.
In this example, the sum of squares due to error (SSE) and the adjusted
R-square statistics are used to help determine the best fit. As described in
“Goodness of Fit Statistics” on page 3-29, the SSE statistic is the least squares
error of the fit, with a value closer to zero indicating a better fit. The adjusted
R-square statistic is generally the best indicator of the fit quality when you add
additional coefficients to your model.
You can modify the information displayed in the Table of Fits with the Table
Options GUI. You open this GUI by clicking the Table options button on the
Fitting GUI. As shown below, select the adjusted R-square statistic and clear
the R-square statistic.

Do not display the R-square

statistic in the Table of Fits.
Display the adjusted R-square
statistic in the Table of Fits.

1-11
1 Getting Started with the Curve Fitting Toolbox

The numerical fit results are shown below. You can click the Table of Fits
column headings to sort by statistics results.
The SSE for exp1 indicates it is a poor fit, which was already determined by
examining the fit and residuals. The lowest SSE value is associated with poly6.
However, the behavior of this fit beyond the data range makes it a poor choice
for extrapolation. The next best SSE value is associated with the fifth degree
polynomial fit, poly5, suggesting it may be the best fit. However, the SSE and
adjusted R-square values for the remaining polynomial fits are all very close to
each other. Which one should you choose?

The confidence bounds for the p1-p3

coefficients suggest that a fifth degree
polynomial overfits the census data.

Click this column heading to sort

the fits by the SSE values.

The SSE and adjusted R-square

values suggest that the fifth degree
polynomial fit is the best one.

1-12
 Fitting the Data

To resolve this issue, examine the confidence bounds for the remaining fits. By
default, 95% confidence bounds are calculated. You can change this level by
selecting the menu item View->Confidence Level from the Curve Fitting Tool.
The p1, p2, and p3 coefficients for the fifth degree polynomial suggest that it
overfits the census data. However, the confidence bounds for the quadratic fit,
poly2, indicate that the fitted coefficients are known fairly accurately.
Therefore, after examining both the graphical and numerical fit results, it
appears that you should use poly2 to extrapolate the census data.

Note The fitted coefficients associated with the constant, linear, and
quadratic terms are nearly identical for each polynomial equation. However,
as the polynomial degree increases, the coefficient bounds associated with the
higher degree terms increase, which suggests overfitting.

For more information about confidence bounds, refer to “Confidence and

Prediction Bounds” on page 3-32.

Saving the Fit Results

By clicking the Save to workspace button, you can save the selected fit and
the associated fit results to the MATLAB workspace. The fit is saved as a
MATLAB object and the associated fit results are saved as structures. This
example saves all the fit results for the best fit, poly2.

fittedmodel1 is saved as a Curve Fitting Toolbox cfit object.

whos fittedmodel1

Name Size Bytes Class

fittedmodel1 1x1 6178 cfit object

Grand total is 386 elements using 6178 bytes

1-13
1 Getting Started with the Curve Fitting Toolbox

The cfit object display includes the model, the fitted coefficients, and the
confidence bounds for the fitted coefficients.
fittedmodel1

fittedmodel1 =
Linear model Poly2:
fittedmodel1(x) = p1*x^2 + p2*x + p3
Coefficients (with 95% confidence bounds):
p1 = 0.006541 (0.006124, 0.006958)
p2 = -23.51 (-25.09, -21.93)
p3 = 2.113e+004 (1.964e+004, 2.262e+004)

The goodness1 structure contains goodness of fit results.

goodness1

goodness1 =
sse: 159.0293
rsquare: 0.9987
dfe: 18
adjrsquare: 0.9986
rmse: 2.9724

The output1 structure contains additional information associated with the fit.
output1

output1 =
numobs: 21
numparam: 3
residuals: [21x1 double]
Jacobian: [21x3 double]
exitflag: 1
algorithm: 'QR factorization and solve'

1-14
 Analyzing the Fit

Analyzing the Fit

You can evaluate (interpolate or extrapolate), differentiate, or integrate a fit
over a specified data range with the Analysis GUI. You open this GUI by
clicking the Analysis button on the Curve Fitting Tool.
For this example, you will extrapolate the quadratic polynomial fit to predict
the US population from the year 2000 to the year 2050 in 10 year increments,
and then plot both the analysis results and the data. To do this:

• Enter the appropriate MATLAB vector in the Analyze at Xi field.

• Select the Evaluate fit at Xi check box.
• Select the Plot results and Plot data set check boxes.
• Click the Apply button.

The numerical extrapolation results are shown below.

Specify the fit and

data to analyze.

Select this check box

to extrapolate.

Plot both the analysis

results and the data.

1-15
1 Getting Started with the Curve Fitting Toolbox

The extrapolated values and the census data set are displayed together in a
new figure window.

Saving the Analysis Results

By clicking the Save to workspace button, you can save the extrapolated
values as a structure to the MATLAB workspace.

The resulting structure is shown below.

analysisresults1

analysisresults1 =
xi: [6x1 double]
yfit: [6x1 double]

1-16
 Saving Your Work

Saving Your Work

The Curve Fitting Toolbox provides you with several options for saving your
work. For example, as described in “Saving the Fit Results” on page 1-13, you
can save one or more fits and the associated fit results as variables to the
MATLAB workspace. You can then use this saved information for
documentation purposes, or to extend your data exploration and analysis. In
addition to saving your work to MATLAB workspace variables, you can
• Save the session.
• Generate an M-file.

Before performing any of these tasks, you may want to remove unwanted data
sets and fits from the Curve Fitting Tool display. An easy way to do this is with
the Plotting GUI. The Plotting GUI shown below is configured to display only
the census data and the best fit, poly2.

Clear the remaining fits

associated with the census
data except the best fit.

1-17
1 Getting Started with the Curve Fitting Toolbox

Saving the Session

The curve fitting session is defined as the current collection of fits for all data
sets. You may want to save your session so that you can continue data
exploration and analysis at a later time using the Curve Fitting Tool without
losing any current work.
Save the current curve fitting session by selecting the menu item File->Save
Session from the Curve Fitting Tool. The Save Session dialog is shown below.

The session is stored in binary form in a cfit file, and contains this
information:

• All data sets and associated fits

• The state of the Fitting GUI, including Table of Fits entries and exclusion
rules
• The state of the Plotting GUI

To avoid saving unwanted data sets, you should delete them from the Curve
Fitting Tool. You delete data sets using the Data Sets pane of the Data GUI. If
there are fits associated with the unwanted data sets, they are deleted as well.
You can load a saved session by selecting the menu item File->Load Session
from the Curve Fitting Tool. When the session is loaded, the saved state of the
Curve Fitting Tool display is reproduced, and may display the data, fits,
residuals, and so on. If you open the Fitting GUI, then the loaded fits are
displayed in the Table of Fits. Select a fit from this table to continue your curve
fitting session.

1-18
 Saving Your Work

Generating an M-File
You may want to generate an M-file so that you can continue data exploration
and analysis from the MATLAB command line. You can run the M-file without
modification to recreate the fits and results that you created with the Curve
Fitting Tool, or you can edit and modify the file as needed. For detailed
descriptions of the functions provided by the toolbox, refer to Chapter 4,
“Function Reference.”
If you have many data sets to fit and you want to automate the fitting process,
you should use the Curve Fitting Tool to select the appropriate model and fit
options, generate an M-file, and then run the M-file in batch mode.
Save your work to an M-file by selecting the menu item File->Save M-file from
the Curve Fitting Tool. The Save M-File dialog is shown below.

The M-file can capture this information from the Curve Fitting Tool:

• All data set variable names, associated fits, and residuals

• Fit options such as whether the data should be normalized, the fit starting
points, and the fitting method

You can recreate the saved fits in a new figure window by typing the name of
the M-file at the MATLAB command line. Note that you must provide the
appropriate data variables as inputs to the M-file. These variables are given in
the M-file help.

1-19
1 Getting Started with the Curve Fitting Toolbox

For example, the help for the censusfit M-file indicates that the variables
cdate and pop are required to recreate the saved fit.
help censusfit

CENSUSFIT Create plot of datasets and fits

CENSUSFIT(CDATE,POP)
Creates a plot, similar to the plot in the main curve fitting
window, using the data that you provide as input. You can
apply this function to the same data you used with cftool
or with different data. You may want to edit the function to
customize the code and this help message.

Number of datasets: 1
Number of fits: 6

1-20
 2
Importing, Viewing, and
Preprocessing Data

This chapter describes how to import, view, and preprocess data with the Curve Fitting Toolbox. You
import data with the Data GUI, and view data graphically as a scatter plot using the Curve Fitting
Tool. The main preprocessing steps are smoothing, and excluding and sectioning data. You smooth
data with the Data GUI, and exclude and section data with the Exclude GUI. The sections are as
follows.

Importing Data Sets Select workspace variables that compose the data set, list all imported
(p. 2-2) and generated data sets, and delete one or more data sets.
Viewing Data (p. 2-6) View the data graphically as a scatter plot.
Smoothing Data (p. 2-9) Reduce noise in a data set using moving average filtering, lowess or
robust lowess, loess or robust loess, or Savitzky-Golay filtering.
Excluding and Sectioning Mark individual data points (outliers) to be excluded from a fit, or mark
Data (p. 2-25) a range of data points (sectioning) to be excluded from a fit.
Additional Preprocessing Additional preprocessing steps not available through the Data GUI,
Steps (p. 2-40) such as transforming the response data and removing Infs, NaNs, and
outliers from a data set.
Selected Bibliography Resources for additional information.
(p. 2-42)
2 Importing, Viewing, and Preprocessing Data

Importing Data Sets

You import data sets into the Curve Fitting Tool with the Data Sets pane of the
Data GUI. Using this pane, you can

• Select workspace variables that compose a data set

• Display a list of all imported data sets
• View, delete, or rename one or more data sets

The Data Sets pane is shown below followed by a description of its features.

Construct and
name the data set.

Data sets list

2-2
 Importing Data Sets

Construct and Name the Data Set

• Import workspace vectors — All selected variables must be the same
length. You can import only vectors, not matrices or scalars. Infs and NaNs
are ignored because you cannot fit data containing these values, and only the
real part of a complex number is used. To perform any curve-fitting task, you
must select at least one vector of data:
- X data — Select the predictor data.
- Y data — Select the response data.
- Weights — Select the weights associated with the response data. If
weights are not imported, they are assumed to be 1 for all data points.
• Preview — The selected workspace vectors are displayed graphically in the
preview window. Weights are not displayed.
• Data set name — The name of the imported data set. The toolbox
automatically creates a unique name for each imported data set. You can
change the name by editing this field. Click the Create data set button to
complete the data import process.

Data Sets List

• Data sets — Lists all data sets added to the Curve Fitting Tool. The data sets
can be created from workspace variables, or from smoothing an existing
imported data set. When you select a data set, you can perform these actions:
- Click View to open the View Data Set GUI. Using this GUI, you can view
a single data set both graphically and numerically. Additionally, you can
display data points to be excluded in a fit by selecting an exclusion rule.
- Click Rename to change the name of a single data set.
- Click Delete to delete one or more data sets. To select multiple data sets,
you can use the Ctrl key and the mouse to select data sets one by one, or
you can use the Shift key and the mouse to select a range of data sets.

2-3
2 Importing, Viewing, and Preprocessing Data

Example: Importing Data

This example imports the ENSO data set into the Curve Fitting Toolbox using
the Data Sets pane of the Data GUI. The first step is to load the data from the
file enso.mat into the MATLAB workspace.
load enso

The workspace contains two new variables, pressure and month:

• pressure is the monthly averaged atmospheric pressure differences between

Easter Island and Darwin, Australia. This difference drives the trade winds
in the southern hemisphere.
• month is the relative time in months.

Alternatively, you can import data by specifying the variable names as

arguments to the cftool function.
cftool(month,pressure)

In this case, the Data GUI is not opened.

Data Import Process

The data import process is described below:

1 Select workspace variables.

The predictor and response data are displayed graphically in the Preview
window. Weights and data points containing Infs or NaNs are not displayed.

2 Specify the data set name.

You should specify a meaningful name when you import multiple data sets.
If you do not specify a name, the default name, which is constructed from the
selected variable names, is used.

2-4
 Importing Data Sets

3 Click the Create data set button.

The Data sets list box displays all the data sets added to the toolbox. Note
that you can construct data sets from workspace variables, or by smoothing
an existing data set.

If your data contains Infs or complex values, a warning message such as the
message shown below is displayed.

The Data Sets pane shown below displays the imported ENSO data in the
Preview window. After you click the Create data set button, the data set enso
is added to the Data sets list box. You can then view, rename, or delete enso by
selecting it in the list box and clicking the appropriate button.

Select the workspace

variable names.

Specify the data set name.

Click Create data set to
import the data.

2-5
2 Importing, Viewing, and Preprocessing Data

Viewing Data
The Curve Fitting Toolbox provides two ways to view imported data:

• Graphically in a scatter plot

• Numerically in a table

Viewing Data Graphically

After you import a data set, it is automatically displayed as a scatter plot in the
Curve Fitting Tool. The response data is plotted on the vertical axis and the
predictor data is plotted on the horizontal axis.
The scatter plot is a powerful tool because it allows you to view the entire data
set at once, and it can easily display a wide range of relationships between the
two variables. You should examine the data carefully to determine whether
preprocessing is required, or to deduce a reasonable fitting approach. For
example, it’s typically very easy to identify outliers in a scatter plot, and to
determine whether you should fit the data with a straight line, a periodic
function, a sum of Gaussians, and so on.

Enhancing the Graphical Display

The Curve Fitting Toolbox provides several tools for enhancing the graphical
display of a data set. These tools are available through the Tools menu, the
GUI toolbar, and right-click menus.
You can zoom in, zoom out, display data and fit tips, and so on using the Tools
menu and the GUI toolbar shown below.

GUI toolbar

Tools menu

2-6
 Viewing Data

You can change the color, line width, line style, and marker type of the
displayed data points using the right-click menu shown below. You activate
this menu by placing your mouse over a data point and right-clicking. Note that
a similar menu is available for fitted curves.

Right-click menu

The ENSO data is shown below after the display has been enhanced using
several of these tools.

Display the legend for

the ENSO data set.

Display data tips for the

maximum response value.

Change the color, marker

type and line style for
the data.

Display the grid.

Change the axis limits.

2-7
2 Importing, Viewing, and Preprocessing Data

Viewing Data Numerically

You can view the numerical values of a data set, as well as data points to be
excluded from subsequent fits, with the View Data Set GUI. You open this GUI
by selecting a name in the Data sets list box of the Data GUI and clicking the
View button. The View Data Set GUI for the ENSO data set is shown below,
followed by a description of its features.

• Data set — Lists the names of the viewed data set and the associated
variables. The data is displayed graphically below this list.
The index, predictor data (X), response data (Y), and weights (if imported)
are displayed numerically in the table. If the data contains Infs or NaNs,
those values are labeled “ignored.” If the data contains complex numbers,
only the real part is displayed.
• Exclusion rules — Lists all the exclusion rules that are compatible with the
viewed data set. When you select an exclusion rule, the data points marked
for exclusion are grayed in the table, and are identified with an “x” in the
graphical display. To exclude the data points while fitting, you must select
the exclusion rule in the Fitting GUI.
An exclusion rule is compatible with the viewed data set if their lengths are
the same, or if it is created by sectioning only.

2-8
 Smoothing Data

Smoothing Data
If your data is noisy, you might need to apply a smoothing algorithm to expose
its features, and to provide a reasonable starting approach for parametric
fitting. The two basic assumptions that underlie smoothing are

• The relationship between the response data and the predictor data is
smooth.
• The smoothing process results in a smoothed value that is a better estimate
of the original value because the noise has been reduced.
The smoothing process attempts to estimate the average of the distribution
of each response value. The estimation is based on a specified number of
neighboring response values.

You can think of smoothing as a local fit because a new response value is
created for each original response value. Therefore, smoothing is similar to
some of the nonparametric fit types supported by the toolbox, such as
smoothing spline and cubic interpolation. However, this type of fitting is not
the same as parametric fitting, which results in a global parameterization of
the data.

Note You should not fit data with a parametric model after smoothing,
because the act of smoothing invalidates the assumption that the errors are
normally distributed. Instead, you should consider smoothing to be a data
exploration technique.

There are two common types of smoothing methods: filtering (averaging) and
local regression. Each smoothing method requires a span. The span defines a
window of neighboring points to include in the smoothing calculation for each
data point. This window moves across the data set as the smoothed response
value is calculated for each predictor value. A large span increases the
smoothness but decreases the resolution of the smoothed data set, while a
small span decreases the smoothness but increases the resolution of the
smoothed data set. The optimal span value depends on your data set and the
smoothing method, and usually requires some experimentation to find.

2-9
2 Importing, Viewing, and Preprocessing Data

The Curve Fitting Toolbox supports these smoothing methods:

• Moving average filtering — Lowpass filter that takes the average of
neighboring data points.
• Lowess and loess — Locally weighted scatter plot smooth. These methods
use linear least squares fitting, and a first-degree polynomial (lowess) or a
second-degree polynomial (loess). Robust lowess and loess methods that are
resistant to outliers are also available.

• Savitzky-Golay filtering — A generalized moving average where you derive

the filter coefficients by performing an unweighted linear least squares fit
using a polynomial of the specified degree.

Note that you can also smooth data using a smoothing spline. Refer to
“Nonparametric Fitting” on page 3-68 for more information.
You smooth data with the Smooth pane of the Data GUI. The pane is shown
below followed by a description of its features.

Data sets

Smoothing method
and parameters

Data sets list

2-10
 Smoothing Data

Data Sets
• Original data set — Select the data set you want to smooth.
• Smoothed data set — Specify the name of the smoothed data set. Note that
the process of smoothing the original data set always produces a new data
set containing smoothed response values.

Smoothing Method and Parameters

• Method — Select the smoothing method. Each response value is replaced
with a smoothed value that is calculated by the specified smoothing method.
- Moving average — Filter the data by calculating an average.
- Lowess — Locally weighted scatter plot smooth using linear least squares
fitting and a first-degree polynomial.
- Loess — Locally weighted scatter plot smooth using linear least squares
fitting and a second-degree polynomial.
- Savitzky-Golay — Filter the data with an unweighted linear least
squares fit using a polynomial of the specified degree.
- Robust Lowess — Lowess method that is resistant to outliers.
- Robust Loess — Loess method that is resistant to outliers.
• Span — The number of data points used to compute each smoothed value.
For the moving average and Savitzky-Golay methods, the span must be odd.
For all locally weighted smoothing methods, if the span is less than 1, it is
interpreted as the percentage of the total number of data points.
• Degree — The degree of the polynomial used in the Savitzky-Golay method.
The degree must be smaller than the span.

Data Sets List

• Smoothed data sets — Lists all the smoothed data sets. You add a smoothed
data set to the list by clicking the Create smoothed data set button. When
you select a data set from the list, you can perform these actions:
- Click View to open the View Data Set GUI. Using this GUI, you can view
a single data set both graphically and numerically. Additionally, you can
display data points to be excluded in a fit by selecting an exclusion rule.
- Click Rename to change the name of a single data set.

2-11
2 Importing, Viewing, and Preprocessing Data

- Click Delete to delete one or more data sets. To select multiple data sets,
you can use the Ctrl key and the mouse to select data sets one by one, or
you can use the Shift key and the mouse to select a range of data sets.
- Click Save to workspace to save a single data set to a structure.

Moving Average Filtering

A moving average filter smooths data by replacing each data point with the
average of the neighboring data points defined within the span. This process is
equivalent to lowpass filtering with the response of the smoothing given by the
difference equation

1
y s ( i ) = ------------------ ( y ( i + N ) + y ( i + N – 1 ) + … + y ( i – N ) )
2N + 1
where ys(i) is the smoothed value for the ith data point, N is the number of
neighboring data points on either side of ys(i), and 2N+1 is the span.
The moving average smoothing method used by the Curve Fitting Toolbox
follows these rules:

• The span must be odd.

• The data point to be smoothed must be at the center of the span.
• The span is adjusted for data points that cannot accommodate the specified
number of neighbors on either side.
• The end points are not smoothed because a span cannot be defined.

Note that you can use MATLAB’s filter function to implement difference
equations such as the one shown above. However, because of the way that the
end points are treated, the toolbox moving average result will differ from the
result returned by filter. Refer to “Difference Equations and Filtering” in the
MATLAB documentation for more information.
For example, suppose you smooth data using a moving average filter with a
span of 5. Using the rules described above, the first four elements of ys are
given by
ys(1) = y(1)
ys(2) = (y(1)+y(2)+y(3))/3
ys(3) = (y(1)+y(2)+y(3)+y(4)+y(5))/5
ys(4) = (y(2)+y(3)+y(4)+y(5)+y(6))/5

2-12
 Smoothing Data

Note that ys(1), ys(2), ... ,ys(end) refer to the order of the data after sorting,
and not necessarily the original order.
The smoothed values and spans for the first four data points of a generated
data set are shown below.

Moving Average Smoothing

80 80
Data Data
Smoothed value Smoothed value
60 60

40 40

20 20

0 0
0 2 4 6 8 0 2 4 6 8
(a) (b)

80 80
Data Data
Smoothed value Smoothed value
60 60

40 40

20 20

0 0
0 2 4 6 8 0 2 4 6 8
(c) (d)

Plot (a) indicates that the first data point is not smoothed because a span
cannot be constructed. Plot (b) indicates that the second data point is
smoothed using a span of three. Plots (c) and (d) indicate that a span of five
is used to calculate the smoothed value.

2-13
2 Importing, Viewing, and Preprocessing Data

Lowess and Loess: Local Regression Smoothing

The names “lowess” and “loess” are derived from the term “locally weighted
scatter plot smooth,” as both methods use locally weighted linear regression to
smooth data.
The smoothing process is considered local because, like the moving average
method, each smoothed value is determined by neighboring data points defined
within the span. The process is weighted because a regression weight function
is defined for the data points contained within the span. In addition to the
regression weight function, you can use a robust weight function, which makes
the process resistant to outliers. Finally, the methods are differentiated by the
model used in the regression: lowess uses a linear polynomial, while loess uses
a quadratic polynomial.
The local regression smoothing methods used by the Curve Fitting Toolbox
follow these rules:

• The span can be even or odd.

• You can specify the span as a percentage of the total number of data points
in the data set. For example, a span of 0.1 uses 10% of the data points.

The regression smoothing and robust smoothing procedures are described in

detail below.

Local Regression Smoothing Procedure

The local regression smoothing process follows these steps for each data point:

1 Compute the regression weights for each data point in the span. The weights
are given by the tricube function shown below.
x–x 3 3
w i =  1 – -------------i 
 d(x ) 
x is the predictor value associated with the response value to be smoothed,
xi are the nearest neighbors of x as defined by the span, and d(x) is the
distance along the abscissa from x to the most distant predictor value within
the span. The weights have these characteristics:
- The data point to be smoothed has the largest weight and the most
influence on the fit.
- Data points outside the span have zero weight and no influence on the fit.

2-14
 Smoothing Data

2 A weighted linear least squares regression is performed. For lowess, the

regression uses a first degree polynomial. For loess, the regression uses a
second degree polynomial.

3 The smoothed value is given by the weighted regression at the predictor

value of interest.

If the smooth calculation involves the same number of neighboring data points
on either side of the smoothed data point, the weight function is symmetric.
However, if the number of neighboring points is not symmetric about the
smoothed data point, then the weight function is not symmetric. Note that
unlike the moving average smoothing process, the span never changes. For
example, when you smooth the data point with the smallest predictor value,
the shape of the weight function is truncated by one half, the leftmost data
point in the span has the largest weight, and all the neighboring points are to
the right of the smoothed value.
The weight function for an end point and for an interior point is shown below
for a span of 31 data points.

Local Regression Weight Function

1.2

0.8 The weight function for

0.6 the leftmost data point
0.4

0.2

0
0 20 40 60 80 100

1.2

0.8 The weight function for

0.6 an interior data point
0.4

0.2

0
0 20 40 60 80 100

2-15
2 Importing, Viewing, and Preprocessing Data

Using the lowess method with a span of five, the smoothed values and
associated regressions for the first four data points of a generated data set are
shown below.

Lowess Smoothing
80 80
Data Data
Smoothed value Smoothed value
60 60

40 40

20 20

0 0
0 2 4 6 8 0 2 4 6 8
(a) (b)

80 80
Data Data
Smoothed value Smoothed value
60 60

40 40

20 20

0 0
0 2 4 6 8 0 2 4 6 8
(c) (d)

Notice that the span does not change as the smoothing process progresses from
data point to data point. However, depending on the number of nearest
neighbors, the regression weight function might not be symmetric about the
data point to be smoothed. In particular, plots (a) and (b) use an asymmetric
weight function, while plots (c) and (d) use a symmetric weight function.
For the loess method, the graphs would look the same except the smoothed
value would be generated by a second-degree polynomial.

2-16
 Smoothing Data

Robust Smoothing Procedure

If your data contains outliers, the smoothed values can become distorted, and
not reflect the behavior of the bulk of the neighboring data points. To overcome
this problem, you can smooth the data using a robust procedure that is not
influenced by a small fraction of outliers. For a description of outliers, refer to
“Marking Outliers” on page 2-27.
The Curve Fitting Toolbox provides a robust version for both the lowess and
loess smoothing methods. These robust methods include an additional
calculation of robust weights, which is resistant to outliers. The robust
smoothing procedure follows these steps:

1 Calculate the residuals from the smoothing procedure described in the

previous section.

2 Compute the robust weights for each data point in the span. The weights are
given by the bisquare function shown below.
2
 ( 1 – ( r i ⁄ 6MAD ) 2 ) r i < 6MAD
wi = 
 0 r i ≥ 6MAD

ri is the residual of the ith data point produced by the regression smoothing
procedure, and MAD is the median absolute deviation of the residuals:

MAD = median ( r )

The median absolute deviation is a measure of how spread out the residuals
are. If ri is small compared to 6MAD, then the robust weight is close to 1. If
ri is greater than 6MAD, the robust weight is 0 and the associated data point
is excluded from the smooth calculation.

3 Smooth the data again using the robust weights. The final smoothed value
is calculated using both the local regression weight and the robust weight.

4 Repeat the previous two steps for a total of five iterations.

2-17
2 Importing, Viewing, and Preprocessing Data

The smoothing results of the lowess procedure are compared below to the
results of the robust lowess procedure for a generated data set that contains a
single outlier. The span for both procedures is 11 data points.

Robust Lowess Smoothing

10
data
lowess

0
0 1 2 3 4 5 6
(a)
5
residuals

−5
0 1 2 3 4 5 6
(b)
10
data
robust lowess

0
0 1 2 3 4 5 6
(c)

Plot (a) shows that the outlier influences the smoothed value for several
nearest neighbors. Plot (b) suggests that the residual of the outlier is greater
than six median absolute deviations. Therefore, the robust weight is zero for
this data point. Plot (c) shows that the smoothed values neighboring the
outlier reflect the bulk of the data.

2-18
 Smoothing Data

Savitzky-Golay Filtering
Savitzky-Golay filtering can be thought of as a generalized moving average.
You derive the filter coefficients by performing an unweighted linear least
squares fit using a polynomial of a given degree. For this reason, a
Savitzky-Golay filter is also called a digital smoothing polynomial filter or a
least squares smoothing filter. Note that a higher degree polynomial makes it
possible to achieve a high level of smoothing without attenuation of data
features.
The Savitzky-Golay filtering method is often used with frequency data or with
spectroscopic (peak) data. For frequency data, the method is effective at
preserving the high-frequency components of the signal. For spectroscopic
data, the method is effective at preserving higher moments of the peak such as
the line width. By comparison, the moving average filter tends to filter out a
significant portion of the signal’s high-frequency content, and it can only
preserve the lower moments of a peak such as the centroid. However,
Savitzky-Golay filtering can be less successful than a moving average filter at
rejecting noise.
The Savitzky-Golay smoothing method used by the Curve Fitting Toolbox
follows these rules:

• The span must be odd.

• The polynomial degree must be less than the span.
• The data points are not required to have uniform spacing.
Normally, Savitzky-Golay filtering requires uniform spacing of the predictor
data. However, the algorithm provided by the Curve Fitting Toolbox
supports nonuniform spacing. Therefore, you are not required to perform an
additional filtering step to create data with uniform spacing.

The plot shown below displays generated Gaussian data and several attempts
at smoothing using the Savitzky-Golay method. The data is very noisy and the
peak widths vary from broad to narrow. The span is equal to 5% of the number
of data points.

2-19
2 Importing, Viewing, and Preprocessing Data

Savitzky−Golay Smoothing
80
noisy data
60

40

20

0
1 2 3 4 5 6 7 8
(a)
80
data
60 S−G quadratic
40

20

0
1 2 3 4 5 6 7 8
(b)
80
data
60 S−G quartic
40

20

0
1 2 3 4 5 6 7 8
(c)

Plot (a) shows the noisy data. To more easily compare the smoothed results,
plots (b) and (c) show the data without the added noise.
Plot (b) shows the result of smoothing with a quadratic polynomial. Notice
that the method performs poorly for the narrow peaks. Plot (c) shows the
result of smoothing with a quartic polynomial. In general, higher degree
polynomials can more accurately capture the heights and widths of narrow
peaks, but can do poorly at smoothing wider peaks.

2-20
 Smoothing Data

Example: Smoothing Data

This example smooths the ENSO data set using the moving average, lowess,
loess, and Savitzky-Golay methods with the default span. As shown below, the
data appears noisy. Smoothing might help you visualize patterns in the data,
and provide insight toward a reasonable approach for parametric fitting.

Because the data appears

noisy, smoothing might
help uncover its structure.

2-21
2 Importing, Viewing, and Preprocessing Data

The Smooth pane shown below displays all the new data sets generated by
smoothing the original ENSO data set. Whenever you smooth a data set, a new
data set of smoothed values is created. The smoothed data sets are
automatically displayed in the Curve Fitting Tool. You can also display a single
data set graphically and numerically by clicking the View button.

A new data set composed of

smoothed values is created
from the original data set.

All smoothed data

sets are listed here.

Click the View button to

display the selected data set.

The View Data Set GUI

displays the selected data set
graphically and numerically.

2-22
 Smoothing Data

Use the Plotting GUI to display only the data sets of interest. As shown below,
the periodic structure of the ENSO data set becomes apparent when it is
smoothed using a moving average filter with the default span. Not
surprisingly, the uncovered structure is periodic, which suggests that a
reasonable parametric model should include trigonometric functions.

Display only the data set

created with the moving
average method.

The smoothing process

uncovers obvious periodic
structure in the data.

Refer to “General Equation: Fourier Series Fit” on page 3-52 for an example
that fits the ENSO data using a sum of sine and cosine functions.

2-23
2 Importing, Viewing, and Preprocessing Data

Saving the Results

By clicking the Save to workspace button, you can save a smoothed data set
as a structure to the MATLAB workspace. This example saves the moving
average results contained in the enso (ma) data set.

The saved structure contains the original predictor data x and the smoothed
data y.
smootheddata1

smootheddata1 =
x: [168x1 double]
y: [168x1 double]

2-24
 Excluding and Sectioning Data

Excluding and Sectioning Data

If there is justification, you might want to exclude part of a data set from a fit.
Typically, you exclude data so that subsequent fits are not adversely affected.
For example, if you are fitting a parametric model to measured data that has
been corrupted by a faulty sensor, the resulting fit coefficients will be
inaccurate.
The Curve Fitting Toolbox provides two methods to exclude data:

• Marking Outliers — Outliers are defined as individual data points that you
exclude because they are inconsistent with the statistical nature of the bulk
of the data.
• Sectioning — Sectioning excludes a window of response or predictor data.
For example, if many data points in a data set are corrupted by large
systematic errors, you might want to section them out of the fit.

For each of these methods, you must create an exclusion rule, which captures
the range, domain, or index of the data points to be excluded.
To exclude data while fitting, you use the Fitting GUI to associate the
appropriate exclusion rule with the data set to be fit. Refer to “Example: Robust
Fit” on page 3-61 for more information about fitting a data set using an
exclusion rule.

2-25
2 Importing, Viewing, and Preprocessing Data

You mark data to be excluded from a fit with the Exclude GUI, which you open
from the Curve Fitting Tool. The GUI is shown below followed by a description
of its features.

Exclusion rule.

Exclude individual
data points.

Exclude data sections

by domain or range.

Exclusion Rule
• Exclusion rule name — Specify the name of the exclusion rule that
identifies the data points to be excluded from subsequent fits.
• Existing exclusion rules — Lists the names of all exclusion rules created
during the current session. When you select an existing exclusion rule, you
can perform these actions:
- Click Copy to copy the exclusion rule. The exclusions associated with the
original exclusion rule are recreated in the GUI. You can modify these
exclusions and then click Create exclusion rule to save them to the copied
rule.
- Click Rename to change the name of the exclusion rule.
- Click Delete to delete the exclusion rule. To select multiple exclusion
rules, you can use the Ctrl key and the mouse to select exclusion rules one
by one, or you can use the Shift key and the mouse to select a range of
exclusion rules.
- Click View to display the exclusion rule graphically. If a data set is
associated with the exclusion rule, the data is also displayed.

2-26
 Excluding and Sectioning Data

Exclude Individual Data Points

• Select data set — Select the data set from which data points will be marked
as excluded. You must select a data set to exclude individual data points.
• Exclude graphically — Open a GUI that allows you to exclude individual
data points graphically.
Individually excluded data points are marked by an “x” in the GUI, and are
automatically identified in the Check to exclude point table.
• Check to exclude point — Select individual data points to exclude. You can
sort this table by clicking on any of the column headings.

Exclude Data Sections by Domain or Range

• Section — Specify a vertical window, a horizontal window, or a box of data
points to include. The excluded data lie outside these windows. You do not
need to select a data set to create an exclusion rule by sectioning.
- Exclude X — Section the predictor data by specifying the domain outside
of which data is excluded.
- Exclude Y — Section the response data by specifying the range outside of
which data is excluded.

Marking Outliers
Outliers are defined as individual data points that you exclude from a fit
because they are inconsistent with the statistical nature of the bulk of the data,
and will adversely affect the fit results. Outliers are often readily identified by
a scatter plot of response data versus predictor data.
Marking outliers with the Curve Fitting Toolbox follows these rules:

• You must specify a data set before creating an exclusion rule.

In general, you should use the exclusion rule only with the specific data set
it was based on. However, the toolbox does not prevent you from using the
exclusion rule with another data set provided the size is the same.
• Using the Exclude GUI, you can exclude outliers either graphically or
numerically.

As described in “Parametric Fitting” on page 3-4, one of the basic assumptions

underlying curve fitting is that the data is statistical in nature and is described

2-27
2 Importing, Viewing, and Preprocessing Data

by a particular distribution, which is often assumed to be Gaussian. The

statistical nature of the data implies that it contains random variations along
with a deterministic component.
data = deterministic component + random component
However, your data set might contain one or more data points that are
nonstatistical in nature, or are described by a different statistical distribution.
These data points might be easy to identify, or they might be buried in the data
and difficult to identify.
A nonstatistical process can involve the measurement of a physical variable
such as temperature or voltage in which the random variation is negligible
compared to the systematic errors. For example, if your sensor calibration is
inaccurate, the data measured with that sensor will be systematically
inaccurate. In some cases, you might be able to quantify this nonstatistical
data component and correct the data accordingly. However, if the scatter plot
reveals that a handful of response values are far removed from neighboring
response values, these data points are considered outliers and should be
excluded from the fit. Outliers are usually difficult to explain away. For
example, it might be that your sensor experienced a power surge or someone
wrote down the wrong number in a log book.
If you decide there is justification, you should mark outliers to be excluded from
subsequent fits — particularly parametric fits. Removing these data points can
have a dramatic effect on the fit results because the fitting process minimizes
the square of the residuals. If you do not exclude outliers, the resulting fit will
be poor for a large portion of your data. Conversely, if you do exclude the
outliers and choose the appropriate model, the fit results should be reasonable.
Because outliers can have a significant effect on a fit, they are considered
influential data. However, not all influential data points are outliers. For
example, your data set can contain valid data points that are far removed from
the rest of the data. The data is valid because it is well described by the model
used in the fit. The data is influential because its exclusion will dramatically
affect the fit results.

2-28
 Excluding and Sectioning Data

Two types of influential data points are shown below for generated data. Also
shown are cubic polynomial fits and a robust fit that is resistant to outliers.

Influential Data Points

150
data
cubic fit

100
These outliers adversely
affect the fit.
50
0 1 2 3 4 5 6 7 8 9 10
(a)
150
data
cubic fit

100

These data points are

consistent with the model.
50
0 1 2 3 4 5 6 7 8 9 10
(b)
150
data
robust cubic fit

100

50
0 1 2 3 4 5 6 7 8 9 10
(c)

Plot (a) shows that the two influential data points are outliers and adversely
affect the fit. Plot (b) shows that the two influential data points are consistent
with the model and do not adversely affect the fit. Plot (c) shows that a robust
fitting procedure is an acceptable alternative to marking outliers for exclusion.
Robust fitting is described in “Robust Least Squares” on page 3-11.

2-29
2 Importing, Viewing, and Preprocessing Data

Sectioning
Sectioning involves specifying a range of response data or a range of predictor
data to exclude. You might want to section a data set because different parts of
the data set are described by different models or many contiguous data points
are corrupted by noise, large systematic errors, and so on.
Sectioning data with the Curve Fitting Toolbox follows these rules:

• If you are only sectioning data and not excluding individual data points, then
you can create an exclusion rule without specifying a data set name.
Note that you can associate the exclusion rule with any data set provided
that the range or domain of the exclusion rule overlaps with the range or
domain of the data set. This is useful if you have multiple data sets from
which you want to exclude data points using the same range or domain.
• Using the Exclude GUI, you specify a range or domain of data to include. The
excluded data lies outside this specification.
Additionally, you can specify only a single range, domain, or box (range and
domain) of included data points. Therefore, at most, you can define two
vertical strips, two horizontal strips, or a border of excluded data. Refer to
“Example: Excluding and Sectioning Data” on page 2-32 for an example.
To exclude multiple sections of data, you can use the excludedata function
from the MATLAB command line.

2-30
 Excluding and Sectioning Data

Two examples of sectioning by domain are shown below for generated data.

Sectioning Data
400
data
200
linear fit
0 cubic fit
−200
−400
−600
−800
1000
1200
0 2 4 6 8 10 12 14 16 18 20
(a)

400
data
200
linear fit
0 cubic fit
−200
−400
−600
−800
1000
1200
0 2 4 6 8 10 12 14 16 18 20
(b)

Plot (a) shows the data set sectioned by fit type. The left section is fit with a
linear polynomial, while the right section is fit with a cubic polynomial. Plot
(b) shows the data set sectioned by fit type and by valid data. Here, the
rightmost section is not part of any fit because the data is corrupted by noise.
Note that reproducing these plots using the toolbox is a multistep process. For
example, to reproduce plot (a), the steps are

1 Create an exclusion rule by sectioning the predictor data such that the data
points described by the linear polynomial are excluded.

2 Create a different exclusion rule by sectioning the predictor data such that
the data points described by the cubic polynomial are excluded.

3 Fit the data twice (once for each exclusion rule) using the appropriate model.

2-31
2 Importing, Viewing, and Preprocessing Data

Example: Excluding and Sectioning Data

This example modifies the ENSO data set to illustrate excluding and
sectioning data. First, copy the ENSO response data to a new variable and add
two outliers that are far removed from the bulk of the data.
rand('state',0)
yy = pressure;
yy(ceil(length(month)*rand(1))) = mean(pressure)*2.5;
yy(ceil(length(month)*rand(1))) = mean(pressure)*3.0;

Import the variables month and yy as the new data set enso1, and open the
Exclude GUI.
Assume that the first and last eight months of the data set are unreliable, and
should be excluded from subsequent fits. The simplest way to exclude these
data points is to section the predictor data. To do this, specify the range of data
you want to include in the Exclude X outside of field of the Section pane.

Data points outside the

specified domain are
marked for exclusion.

There are two ways to exclude individual data points: using the Check to
exclude point table or graphically. For this example, the simplest way to
exclude the outliers is graphically. To do this, select the data set name and click
the Exclude graphically button, which opens the Select Points for Exclusion
Rule GUI.

Select the data set.

Open the GUI to exclude

data points graphically.

2-32
 Excluding and Sectioning Data

To mark data points for exclusion in the GUI, place the mouse cursor over the
data point and left-click. The excluded data point is marked with a red X. To
include an excluded data point, right-click the data point or select the Include
Them radio button and left-click. Included data points are marked with a blue
circle. To select multiple data points, click the left mouse button and drag the
selection rubber band so that the rubber band box encompasses the desired
data points. Note that the GUI identifies sectioned data with gray strips. You
cannot graphically include sectioned data.
As shown below, the first and last eight months of data are excluded from the
data set by sectioning, and the two outliers are excluded graphically. Note that
the graphically excluded data points are identified in the Check to exclude
point table. If you decide to include an excluded data point using the table, the
graph is automatically updated.

The x’s indicate data

points excluded manually.

The vertical gray strips

indicate data points
sectioned by domain.

If there are fits associated with the data, you can exclude data points based on
the residuals of the fit by selecting the residual data in the Y list.

2-33
2 Importing, Viewing, and Preprocessing Data

The Exclude GUI for this example is shown below.

Individual data points

marked for exclusion.

Data points outside the

specified domain are
marked for exclusion.

To save the exclusion rule, click the Create exclusion rule button. To exclude
the data from a fit, you must select the exclusion rule from the Fitting GUI.
Because the exclusion rule created in this example uses individually excluded
data points, you can use it only with data sets that are the same size as the
ENSO data set.

2-34
 Excluding and Sectioning Data

Viewing the Exclusion Rule

To view the exclusion rule, select an existing exclusion rule name and click the
View button. The View Exclusion Rule GUI shown below displays the modified
ENSO data set and the excluded data points, which are grayed in the table.

The excluded data points

are grayed in the table.

Example: Sectioning Periodic Data

For all parametric equations, the toolbox provides coefficient starting values.
For certain types of data sets such as periodic data containing many periods,
the starting values may not lead to satisfactory results. In this case, sectioning
the data can provide you with improved starting values for the fit.
This example uses generated sine data with noise added. The time vector is
given by t and the amplitude, frequency, and phase constant of the data are
given by the vector cf.
rand('state',0);
t = [0:0.005:1.0]';
cf = [10 16*pi pi/4];
noisysine = cf(1)*(sin(cf(2)*t+cf(3))) + (rand(size(t))-0.5);

Import the variables t and noisysine, and fit the data with a single-term sine
equation. The Fitting GUI, Fit Options GUI, and Curve Fitting Tool are shown
below. To display the fit starting values, click the Fit options button. Note that

2-35
2 Importing, Viewing, and Preprocessing Data

the amplitude starting point is reasonably close to the expected value, but the
frequency and phase constant are not, which produces a poor fit.

The amplitude starting point is reasonably

close to the expected value, but the
frequency and phase constant are not.

2-36
 Excluding and Sectioning Data

To produce a reasonable fit, follow these steps:

1 Create an exclusion rule that includes one or two periods, and excludes the
remaining data.

As shown below, an exclusion rule is created graphically by using the

selection rubber band to exclude all data points outside the first period. The
exclusion rule is named 1Period.

Exclude data
graphically.

Use the selection rubber band

to exclude data points outside
the first period.

2-37
2 Importing, Viewing, and Preprocessing Data

2 Create a new fit using the single-term sine equation with the exclusion rule
1Period applied.

The fit looks reasonable throughout the entire data set. However, because
the global fit was based on a small fraction of data, goodness of fit statistics
will not provide much insight into the fit quality.

Apply exclusion rule to the

single-term exponential fit.

The global fit looks reasonable although

an accurate evaluation of the goodness
of fit is not possible.

2-38
 Excluding and Sectioning Data

3 Fit the entire data set using the fitted coefficient values from the previous
step as starting values.

The Fitting GUI, Fit Options GUI, and Curve Fitting Tool are shown below.
Both the numerical and graphical fit results indicate a reasonable fit.

The coefficient starting values are

given by the previous fit results.

2-39
2 Importing, Viewing, and Preprocessing Data

Additional Preprocessing Steps

Additional preprocessing steps not available through the Curve Fitting
Toolbox GUIs include

• Transforming the response data

• Removing Infs, NaNs, and outliers

Transforming the Response Data

In some circumstances, you might want to transform the response data.
Common transformations include the logarithm ln(y), and power functions
such as y1/2, y-1, and so on. Using these transformations, you can linearize a
nonlinear model, contract response data that spans one or more orders of
magnitude, or simplify a model so that it involves fewer coefficients.

Note You must transform variables at the MATLAB command line, and then
import those variables into the Curve Fitting Toolbox. You cannot transform
variables using any of the graphical user interfaces.

For example, suppose you want to use the following model to fit your data.

1
y = -------------------------------
2
ax + bx + c
If you decide to use the power transform y-1, then the transformed model is
given by
–1 2
y = ax + bx + c
As another example, the equation
bx
y = ae
becomes linear if you take the log transform of both sides.

ln ( y ) = ln ( a ) + bx
You can now use linear least squares fitting procedures.

2-40
 Additional Preprocessing Steps

There are several disadvantages associated with performing transformations:

• For the log transformation, negative response values cannot be processed.

• For all transformations, the basic assumption that the residual variance is
constant is violated. To avoid this problem, you could plot the residuals on
the transformed scale. For the power transformation shown above, the
transformed scale is given by the residuals
–1 –1
ri = yi – ŷ i

Note that the residual plot associated with the Curve Fitting Tool does not
support transformed scales.

Deciding on a particular transformation is not always obvious. However, a

scatter plot will often reveal the best form to use. In practice you can
experiment with various transforms and then plot the residuals from the
command line using the transformed scale. If the errors are reasonable (they
appear random with minimal scatter, and don’t exhibit any systematic
behavior), the transform is a good candidate.

Removing Infs, NaNs, and Outliers

Although the Curve Fitting Toolbox ignores Infs and NaNs when fitting data,
and you can exclude outliers during the fitting process, you might still want to
remove this data from your data set. To do so, you modify the associated data
set variables from the MATLAB command line.
For example, when using toolbox functions such as fit from the command line,
you must supply predictor and response vectors that contain finite numbers. To
remove Infs, you can use the isinf function.
ind = find(isinf(xx));
xx(ind) = [];
yy(ind) = [];

To remove NaNs, you can use the isnan function. For examples that remove
NaNs and outliers from a data set, refer to “Data Preprocessing” in the MATLAB
documentation.

2-41
2 Importing, Viewing, and Preprocessing Data

Selected Bibliography
[1] Cleveland, W.S., “Robust Locally Weighted Regression and Smoothing
Scatterplots,” Journal of the American Statistical Association, Vol. 74, pp.
829-836, 1979.
[2] Cleveland, W.S. and S.J. Devlin, “Locally Weighted Regression: An
Approach to Regression Analysis by Local Fitting,” Journal of the American
Statistical Association, Vol. 83, pp. 596-610, 1988.
[3] Chambers, J., W.S. Cleveland, B. Kleiner, and P. Tukey, Graphical Methods
for Data Analysis, Wadsworth International Group, Belmont, CA, 1983.
[4] Press, W.H., S.A. Teukolsky, W.T. Vetterling, and B.P. Flannery,
Numerical Recipes in C, The Art of Scientific Computing, Cambridge
University Press, Cambridge, England, 1993.
[5] Goodall, C., “A Survey of Smoothing Techniques,” Modern Methods of Data
Analysis, (J. Fox and J.S. Long, eds.), Sage Publications, Newbury Park, CA,
pp. 126-176, 1990.
[6] Hutcheson, M.C., “Trimmed Resistant Weighted Scatterplot Smooth,”
Master’s Thesis, Cornell University, Ithaca, NY, 1995.
[7] Orfanidis, S.J., Introduction to Signal Processing, Prentice-Hall, Englewood
Cliffs, NJ, 1996.

2-42
 3

Fitting Data

Curve fitting refers to fitting curved lines to data. The curved line comes from regression techniques,
a spline calculation, or interpolation. The data can be measured from a sensor, generated from a
simulation, historical, and so on. The goal of curve fitting is to gain insight into your data. The insight
will enable you to improve data acquisition techniques for future experiments, accept or refute a
theoretical model, extract physical meaning from fitted coefficients, and draw conclusions about the
data’s parent population.

This chapter describes how to fit data and evaluate the goodness of fit with the Curve Fitting Toolbox.
The sections are as follows.

The Fitting Process (p. 3-2) The general steps you use when fitting any data set.
Parametric Fitting (p. 3-4) Fit your data using parametric models such as polynomials and
exponentials, specify fit options such as the fitting algorithm and
coefficient starting points, and evaluate the goodness of fit using
graphical and numerical techniques.
Parametric fitting produces coefficients that describe the data globally,
and often have physical meaning.
Nonparametric Fitting Fit your data using nonparametric fit types such as splines and
(p. 3-68) interpolants.
Nonparametric fitting is useful when you want to fit a smooth curve
through your data, and you are not interested in interpreting fitted
coefficients.
Selected Bibliography Resources for additional information.
(p. 3-75)
3 Fitting Data

The Fitting Process

You fit data using the Fitting GUI. To open the Fitting GUI, click the Fitting
button from the Curve Fitting Tool.
The Fitting GUI is shown below for the census data described in “Getting
Started with the Curve Fitting Toolbox” on page 1-1, followed by the general
steps you use when fitting any data set.

1. Select a data set and specify

a fit name.
2. Select an exclusion rule.

3. Select a fit type, select fit

options, fit the data, and
evaluate the goodness of fit.

4. Compare the current fit and data

set to other fits and data sets.

5. Save the selected fit results to

the workspace.

3-2
 The Fitting Process

1 Select a data set and fit name.

- Select the name of the current fit. When you click New fit or Copy fit, a
default fit name is automatically created in the Fit name field. You can
specify a new fit name by editing this field.
- Select the name of the current data set from the Data set list. All imported
and smoothed data sets are listed.

2 Select an exclusion rule.

If you want to exclude data from a fit, select an exclusion rule from the
Exclusion rule list. The list contains only exclusion rules that are
compatible with the current data set. An exclusion rule is compatible with
the current data set if their lengths are identical, or if it is created by
sectioning only.

3 Select a fit type and fit options, fit the data, and evaluate the goodness of fit.

- The fit type can be a library or custom parametric model, a smoothing

spline, or an interpolant.
- Select fit options such as the fitting algorithm, and coefficient starting
points and constraints. Depending on your data and model, accepting the
default fit options often produces an excellent fit.
- Fit the data by clicking the Apply button or by selecting the Immediate
apply check box.
- Examine the fitted curve, residuals, goodness of fit statistics, confidence
bounds, and prediction bounds for the current fit.

4 Compare fits.

- Compare the current fit and data set to previous fits and data sets by
examining the goodness of fit statistics.
- Use the Table Options GUI to modify which goodness of fit statistics are
displayed in the Table of Fits. You can sort the table by clicking on any
column heading.

5 Save the fit results.

If the fit is good, save the results as a structure to the MATLAB workspace.
Otherwise, modify the fit options or select another model.

3-3
3 Fitting Data

Parametric Fitting
Parametric fitting involves finding coefficients (parameters) for one or more
models that you fit to data. The data is assumed to be statistical in nature and
is divided into two components: a deterministic component and a random
component.
data = deterministic component + random component
The deterministic component is given by the fit and the random component is
often described as error associated with the data.
data = fit + error
The fit is given by a model that is a function of the independent (predictor)
variable and one or more coefficients. The error represents random variations
in the data that follow a specific probability distribution (usually Gaussian).
The variations can come from many different sources, but are always present
at some level when you are dealing with measured data. Systematic variations
can also exist, but they can be difficult to quantify.
The fitted coefficients often have physical significance. For example, suppose
you have collected data that corresponds to a single decay mode of a radioactive
nuclide, and you want to find the half-life (T1/2) of the decay. The law of
radioactive decay states that the activity of a radioactive substance decays
exponentially in time. Therefore, the model to use in the fit is given by
– λt
y = y0 e

where y0 is the number of nuclei at time t = 0, and λ is the decay constant.

Therefore, the data can be described by
– λt
data = y 0 e + error

Both y0 and λ are coefficients determined by the fit. Because T1/2 = ln(2)/λ, the
fitted value of the decay constant yields the half-life. However, because the
data contains some error, the deterministic component of the equation cannot
completely describe the variability in the data. Therefore, the coefficients and
half-life calculation will have some uncertainty associated with them. If the
uncertainty is acceptable, then you are done fitting the data. If the uncertainty
is not acceptable, then you might have to take steps to reduce the error and
repeat the data collection process.

3-4
 Parametric Fitting

Basic Assumptions About the Error

When fitting data that contains random variations, there are two important
assumptions that are usually made about the error:
• The error exists only in the response data, and not in the predictor data.
• The errors are random and follow a normal (Gaussian) distribution with zero
mean and constant variance, σ2.

The second assumption is often expressed as

2
error ∼ N ( 0, σ )
The components of this expression are described below.

Normal Distribution
The errors are assumed to be normally distributed because the normal
distribution often provides an adequate approximation to the distribution of
many measured quantities. Although the least squares fitting method does not
assume normally distributed errors when calculating parameter estimates, the
method works best for data that does not contain a large number of random
errors with extreme values. The normal distribution is one of the probability
distributions in which extreme random errors are uncommon. However,
statistical results such as confidence and prediction bounds do require
normally distributed errors for their validity.

Zero Mean
If the mean of the errors is zero, then the errors are purely random. If the mean
is not zero, then it might be that the model is not the right choice for your data,
or the errors are not purely random and contain systematic errors.

Constant Variance
A constant variance in the data implies that the “spread” of errors is constant.
Data that has the same variance is sometimes said to be of equal quality.
The assumption that the random errors have constant variance is not implicit
to weighted least squares regression. Instead, it is assumed that the weights
provided in the fitting procedure correctly indicate the differing levels of
quality present in the data. The weights are then used to adjust the amount of
influence each data point has on the estimates of the fitted coefficients to an
appropriate level.

3-5
3 Fitting Data

The Least Squares Fitting Method

The Curve Fitting Toolbox uses the method of least squares when fitting data.
The fitting process requires a model that relates the response data to the
predictor data with one or more coefficients. The result of the fitting process is
an estimate of the “true” but unknown coefficients of the model.
To obtain the coefficient estimates, the least squares method minimizes the
summed square of residuals. The residual for the ith data point ri is defined as
the difference between the observed response value yi and the fitted response
value ŷ i , and is identified as the error associated with the data.

r i = y i – ŷ i
residual = data – fit
The summed square of residuals is given by
n n
2 2
S = ∑ ri = ∑ ( yi – ŷi )
i=1 i=1

where n is the number of data points included in the fit and S is the sum of
squares error estimate. The supported types of least squares fitting include

• Linear least squares

• Weighted linear least squares
• Robust least squares
• Nonlinear least squares

Linear Least Squares

The Curve Fitting Toolbox uses the linear least squares method to fit a linear
model to data. A linear model is defined as an equation that is linear in the
coefficients. For example, polynomials are linear but Gaussians are not. To
illustrate the linear least squares fitting process, suppose you have n data
points that can be modeled by a first-degree polynomial.

y = p 1 x + p2

3-6
 Parametric Fitting

To solve this equation for the unknown coefficients p1 and p2, you write S as a
system of n simultaneous linear equations in two unknowns. If n is greater
than the number of unknowns, then the system of equations is overdetermined.
n
2
S = ∑ ( y i – ( p 1 xi + p 2 ) )
i=1

Because the least squares fitting process minimizes the summed square of the
residuals, the coefficients are determined by differentiating S with respect to
each parameter, and setting the result equal to zero.
n
∂S
∂ p1
= –2 ∑ x i ( yi – ( p 1 x i + p 2 ) ) = 0
i=1
n
∂S
∂ p2
= –2 ∑ ( yi – ( p1 x i + p 2 ) ) = 0
i=1

The estimates of the true parameters are usually represented by b.

Substituting b1 and b2 for p1 and p2, the previous equations become

∑ xi ( yi – ( b1 xi + b2 ) ) = 0

∑ ( yi – ( b1 xi + b2 ) ) = 0

where the summations run from i =1 to n. The normal equations are defined as
2
b1 ∑ xi + b 2 ∑ x i = ∑ xi y i
b1 ∑ xi + nb 2 = ∑ yi
Solving for b1

n xi yi – ∑ ∑ ∑ xi yi
b 1 = ----------------------------------------------------
2
n x i –  x i
2
∑  ∑ 

3-7
3 Fitting Data

Solving for b2 using the b1 value

1
b 2 = ---  ∑ yi – b1 ∑ xi
n
As you can see, estimating the coefficients p1 and p2 requires only a few simple
calculations. Extending this example to a higher degree polynomial is
straightforward although a bit tedious. All that is required is an additional
normal equation for each linear term added to the model.
In matrix form, linear models are given by the formula

y = Xβ + ε
where

• y is an n-by-1 vector of responses.

• β is a m-by-1 vector of coefficients.
• X is the n-by-m design matrix for the model.
• ε is an n-by-1 vector of errors.

For the first-degree polynomial, the n equations in two unknowns are

expressed in terms of y, X, and β as

y1 x1 1
y2 x2 1
y3 x3 1 p1
= ×
. . p2
. .
. .
yn xn 1

The least squares solution to the problem is a vector b, which estimates the
unknown vector of coefficients β. The normal equations are given by
T T
( X X )b = X y

3-8
 Parametric Fitting

where XT is the transpose of the design matrix X. Solving for b,

T –1 T
b = (X X) X y
In MATLAB, you can use the backslash operator to solve a system of
simultaneous linear equations for unknown coefficients. Because inverting
XTX can lead to unacceptable rounding errors, MATLAB uses QR
decomposition with pivoting, which is a very stable algorithm numerically.
Refer to “Arithmetic Operators” in the MATLAB documentation for more
information about the backslash operator and QR decomposition.
You can plug b back into the model formula to get the predicted response
values, ŷ .

ŷ = Xb = Hy
T –1 T
H = X (X X) X
A hat (circumflex) over a letter denotes an estimate of a parameter or a
prediction from a model. The projection matrix H is called the hat matrix,
because it puts the hat on y.
The residuals are given by

r = y – ŷ = ( 1 – H )y
Refer to [1] or [2] for a complete description of the matrix representation of
least squares regression.

Weighted Linear Least Squares

As described in “Basic Assumptions About the Error” on page 3-5, it is usually
assumed that the response data is of equal quality and, therefore, has constant
variance. If this assumption is violated, your fit might be unduly influenced by
data of poor quality. To improve the fit, you can use weighted least squares
regression where an additional scale factor (the weight) is included in the
fitting process. Weighted least squares regression minimizes the error
estimate
n
2
S = ∑ wi ( yi – ŷi )
i=1

3-9
3 Fitting Data

where wi are the weights. The weights determine how much each response
value influences the final parameter estimates. A high-quality data point
influences the fit more than a low-quality data point. Weighting your data is
recommended if the weights are known, or if there is justification that they
follow a particular form.
The weights modify the expression for the parameter estimates b in the
following way,

T –1 T
b = β̂ = ( X WX ) X Wy

where W is given by the diagonal elements of the weight matrix w.

You can often determine whether the variances are not constant by fitting the
data and plotting the residuals. In the plot shown below, the data contains
replicate data of various quality and the fit is assumed to be correct. The poor
quality data is revealed in the plot of residuals, which has a “funnel” shape
where small predictor values yield a bigger scatter in the response values than
large predictor values.
100
data
fitted curve
80

60
y

40

20

0
0 1 2 3 4 5 6
x

15
residuals
10

−5

−10

−15
0 1 2 3 4 5 6

3-10
 Parametric Fitting

The weights you supply should transform the response variances to a constant
value. If you know the variances of your data, then the weights are given by
2
wi = 1 ⁄ σ

If you don’t know the variances, you can approximate the weights using an
equation such as
n –1
 
1
w i =  --- ( yi – y ) 
2
n ∑ 
 i=1 

This equation works well if your data set contains replicates. In this case, n is
the number of sets of replicates. However, the weights can vary greatly. A
better approach might be to plot the variances and fit the data using a sensible
model. The form of the model is not very important — a polynomial or power
function works well in many cases.

Robust Least Squares

As described in “Basic Assumptions About the Error” on page 3-5, it is usually
assumed that the response errors follow a normal distribution, and that
extreme values are rare. Still, extreme values called outliers do occur.
The main disadvantage of least squares fitting is its sensitivity to outliers.
Outliers have a large influence on the fit because squaring the residuals
magnifies the effects of these extreme data points. To minimize the influence
of outliers, you can fit your data using robust least squares regression. The
toolbox provides these two robust regression schemes:

• Least absolute residuals (LAR) — The LAR scheme finds a curve that
minimizes the absolute difference of the residuals, rather than the squared
differences. Therefore, extreme values have a lesser influence on the fit.
• Bisquare weights — This scheme minimizes a weighted sum of squares,
where the weight given to each data point depends on how far the point is
from the fitted line. Points near the line get full weight. Points farther from

3-11
3 Fitting Data

the line get reduced weight. Points that are farther from the line than would
be expected by random chance get zero weight.
For most cases, the bisquare weight scheme is preferred over LAR because it
simultaneously seeks to find a curve that fits the bulk of the data using the
usual least squares approach, and it minimizes the effect of outliers.

Robust fitting with bisquare weights uses an iteratively reweighted least

squares algorithm, and follows this procedure:

1 Fit the model by weighted least squares.

2 Compute the adjusted residuals and standardize them. The adjusted

residuals are given by
ri
r adj = -------------------
1 – hi

ri are the usual least squares residuals and hi are leverages that adjust the
residuals by downweighting high-leverage data points, which have a large
effect on the least squares fit. The standardized adjusted residuals are given
by

r adj
u = ----------
Ks

K is a tuning constant equal to 4.685, and s is the robust variance given by

MAD/0.6745 where MAD is the median absolute deviation of the residuals.
Refer to [7] for a detailed description of h, K, and s.

3 Compute the robust weights as a function of u. The bisquare weights are

given by
2
 ( 1 – ( ui )2 ) ui < 1
wi = 
 0 ui ≥ 1

Note that if you supply your own regression weight vector, the final weight
is the product of the robust weight and the regression weight.

3-12
 Parametric Fitting

4 If the fit converges, then you are done. Otherwise, perform the next iteration
of the fitting procedure by returning to the first step.

The plot shown below compares a regular linear fit with a robust fit using
bisquare weights. Notice that the robust fit follows the bulk of the data and is
not strongly influenced by the outliers.

30
Data
Regular linear fit
Robust fit w/bisquare weights
25

20

15
y

10

−5
0 2 4 6 8 10 12 14 16 18 20
x

Instead of minimizing the effects of outliers by using robust regression, you can
mark data points to be excluded from the fit. Refer to “Excluding and
Sectioning Data” on page 2-25 for more information.

3-13
3 Fitting Data

Nonlinear Least Squares

The Curve Fitting Toolbox uses the nonlinear least squares formulation to fit
a nonlinear model to data. A nonlinear model is defined as an equation that is
nonlinear in the coefficients, or a combination of linear and nonlinear in the
coefficients. For example, Gaussians, ratios of polynomials, and power
functions are all nonlinear.
In matrix form, nonlinear models are given by the formula

y = f ( X, β ) + ε
where

• y is an n-by-1 vector of responses.

• f is a function of β and X.
• β is a m-by-1 vector of coefficients.
• X is the n-by-m design matrix for the model.
• ε is an n-by-1 vector of errors.

Nonlinear models are more difficult to fit than linear models because the
coefficients cannot be estimated using simple matrix techniques. Instead, an
iterative approach is required that follows these steps:

1 Start with an initial estimate for each coefficient. For some nonlinear
models, a heuristic approach is provided that produces reasonable starting
values. For other models, random values on the interval [0,1] are provided.

2 Produce the fitted curve for the current set of coefficients. The fitted
response value ŷ is given by

ŷ = f ( X, b )
and involves the calculation of the Jacobian of f(X,b), which is defined as a
matrix of partial derivatives taken with respect to the coefficients.

3-14
 Parametric Fitting

3 Adjust the coefficients and determine whether the fit improves. The
direction and magnitude of the adjustment depend on the fitting algorithm.
The toolbox provides these algorithms:
- Trust-region — This is the default algorithm and must be used if you
specify coefficient constraints. It can solve difficult nonlinear problems
more efficiently than the other algorithms and it represents an
improvement over the popular Levenberg-Marquardt algorithm.
- Levenberg-Marquardt — This algorithm has been used for many years
and has proved to work most of the time for a wide range of nonlinear
models and starting values. If the trust-region algorithm does not produce
a reasonable fit, and you do not have coefficient constraints, you should try
the Levenberg-Marquardt algorithm.
- Gauss-Newton — This algorithm is potentially faster than the other
algorithms, but it assumes that the residuals are close to zero. It’s included
with the toolbox for pedagogical reasons and should be the last choice for
most models and data sets.

For more information about the trust region algorithm, refer to [4] and to
“Trust Region Methods for Nonlinear Minimization” in the Optimization
Toolbox documentation. For more information about the
Levenberg-Marquardt and Gauss-Newton algorithms, refer to “Nonlinear
Least Squares Implementation” in the same guide. Additionally, the
Levenberg-Marquardt algorithm is described in [5] and [6].

4 Iterate the process by returning to step 2 until the fit reaches the specified
convergence criteria.

You can use weights and robust fitting for nonlinear models, and the fitting
process is modified accordingly.
Because of the nature of the approximation process, no algorithm is foolproof
for all nonlinear models, data sets, and starting points. Therefore, if you do not
achieve a reasonable fit using the default starting points, algorithm, and
convergence criteria, you should experiment with different options. Refer to
“Specifying Fit Options” on page 3-23 for a description of how to modify the
default options. Because nonlinear models can be particularly sensitive to the
starting points, this should be the first fit option you modify.

3-15
3 Fitting Data

Library Models
The parametric library models provided by the Curve Fitting Toolbox are
described below.

Exponentials
The toolbox provides a one-term and a two-term exponential model.
bx
y = ae
bx dx
y = ae + ce
Exponentials are often used when the rate of change of a quantity is
proportional to the initial amount of the quantity. If the coefficient associated
with e is negative, y represents exponential decay. If the coefficient is positive,
y represents exponential growth.
For example, a single radioactive decay mode of a nuclide is described by a
one-term exponential. a is interpreted as the initial number of nuclei, b is the
decay constant, x is time, and y is the number of remaining nuclei after a
specific amount of time passes. If two decay modes exist, then you must use the
two-term exponential model. For each additional decay mode, you add another
exponential term to the model.
Examples of exponential growth include contagious diseases for which a cure
is unavailable, and biological populations whose growth is uninhibited by
predation, environmental factors, and so on.

Fourier Series
The Fourier series is a sum of sine and cosine functions that is used to describe
a periodic signal. It is represented in either the trigonometric form or the
exponential form. The toolbox provides the trigonometric Fourier series form
shown below,
n

y = a0 + ∑ ai cos ( nwx ) + bi sin ( nwx )

i=1

where a0 models any DC offset in the signal and is associated with the i = 0
cosine term, w is the fundamental frequency of the signal, n is the number of
terms (harmonics) in the series, and 1 ≤ n ≤ 8 .

3-16
 Parametric Fitting

For more information about the Fourier series, refer to “Fourier Analysis and
the Fast Fourier Transform” in the MATLAB documentation. For an example
that fits the ENSO data to a custom Fourier series model, refer to “General
Equation: Fourier Series Fit” on page 3-52.

Gaussian
The Gaussian model is used for fitting peaks, and is given by the equation

n x – bi 2
–  -------------
ci
y= ∑ ai e
i=1

where a is the amplitude, b is the centroid (location), c is related to the peak

width, n is the number of peaks to fit, and 1 ≤ n ≤ 8 .
Gaussian peaks are encountered in many areas of science and engineering. For
example, line emission spectra and chemical concentration assays can be
described by Gaussian peaks. For an example that fits two Gaussian peaks and
an exponential background, refer to “General Equation: Gaussian Fit with
Exponential Background” on page 3-57.

Polynomials
Polynomial models are given by
n+1

∑ pi x
n+1–i
y =
i=1

where n + 1 is the order of the polynomial, n is the degree of the polynomial,

and 1 ≤ n ≤ 9 . The order gives the number of coefficients to be fit, and the
degree gives the highest power of the predictor variable.
In this guide, polynomials are described in terms of their degree. For example,
a third-degree (cubic) polynomial is given by
3 2
y = p 1 x + p 2 x + p 3 x + p4

Polynomials are often used when a simple empirical model is required. The
model can be used for interpolation or extrapolation, or it can be used to
characterize data using a global fit. For example, the temperature-to-voltage

3-17
3 Fitting Data

conversion for a Type J thermocouple in the 0o to 760o temperature range is

described by a seventh-degree polynomial.

Note If you do not require a global parametric fit and want to maximize the
flexibility of the fit, piecewise polynomials might provide the best approach.
Refer to “Nonparametric Fitting” on page 3-68 for more information.

The main advantages of polynomial fits include reasonable flexibility for data
that is not too complicated, and they are linear, which means the fitting process
is simple. The main disadvantage is that high-degree fits can become unstable.
Additionally, polynomials of any degree can provide a good fit within the data
range, but can diverge wildly outside that range. Therefore, you should
exercise caution when extrapolating with polynomials. Refer to “Determining
the Best Fit” on page 1-8 for examples of good and poor polynomial fits to
census data.
Note that when you fit with high-degree polynomials, the fitting procedure
uses the predictor values as the basis for a matrix with very large values, which
can result in scaling problems. To deal with this, you should normalize the data
by centering it at zero mean and scaling it to unit standard deviation. You
normalize data by selecting the Center and scale X data check box on the
Fitting GUI.

Power Series
The toolbox provides a one-term and a two-term power series model.
b
y = ax
c
y = a + bx
Power series models are used to describe a variety of data. For example, the
rate at which reactants are consumed in a chemical reaction is generally
proportional to the concentration of the reactant raised to some power.

3-18
 Parametric Fitting

Rationals
Rational models are defined as ratios of polynomials and are given by
n+1

∑ pi x
n+1–i

i=1
y = -------------------------------------------
m
-
m m–i
x + ∑ qi x
i=1

where n is the degree of the numerator polynomial and 0 ≤ n ≤ 5 , while m is the

degree of the denominator polynomial and 1 ≤ m ≤ 5 . Note that the coefficient
m
associated with x is always 1. This makes the numerator and denominator
unique when the polynomial degrees are the same.
In this guide, rationals are described in terms of the degree of the
numerator/the degree of the denominator. For example, a quadratic/cubic
rational equation is given by
2
p1 x + p2 x + p3
y = -----------------------------------------------------
3 2
x + q 1 x + q 2 x + q3

Like polynomials, rationals are often used when a simple empirical model is
required. The main advantage of rationals is their flexibility with data that has
complicated structure. The main disadvantage is that they become unstable
when the denominator is around zero. For an example that uses rational
polynomials of various degrees, refer to “Example: Rational Fit” on page 3-41.

Sum of Sines
The sum of sines model is used for fitting periodic functions, and is given by the
equation
n

y = ∑ a i sin ( bi x + ci )
i=1

where a is the amplitude, b is the frequency, and c is the phase constant for
each sine wave term. n is the number of terms in the series and 1 ≤ n ≤ 8 . This
equation is closely related to the Fourier series described previously. The main

3-19
3 Fitting Data

difference is that the sum of sines equation includes the phase constant, and
does not include a DC offset term.

Weibull Distribution
The Weibull distribution is widely used in reliability and life (failure rate) data
analysis. The toolbox provides the two-parameter Weibull distribution
b
b – 1 – ax
y = abx e
where a is the scale parameter and b is the shape parameter. Note that there
is also a three-parameter Weibull distribution with x replaced by x – c where c
is the location parameter. Additionally, there is a one-parameter Weibull
distribution where the shape parameter is fixed and only the scale parameter
is fitted. To use these distributions, you must create a custom equation.
Note that the Curve Fitting Toolbox does not fit Weibull probability
distributions to a sample of data. Instead, it fits curves to response and
predictor data such that the curve has the same shape as a Weibull
distribution.

Custom Equations
If the toolbox library does not contain the desired parametric equation, you
must create your own custom equation. However, if possible, you should use
the library equations because they offer the best chance for rapid convergence.
This is because

• For most models, optimal default coefficient starting points are calculated.
For custom equations, the default starting points are chosen at random on
the interval [0,1]. Refer to “Default Coefficient Parameters” on page 3-26 for
more information.
• An analytic Jacobian is used instead of finite differencing.
• When using the Analysis GUI, analytic derivatives are calculated as well as
analytic integrals if the integral can be expressed in closed form.

Note To save custom equations for later use, you should save the
curve-fitting session with the File-> Save Session menu item.

3-20
 Parametric Fitting

You create custom equations with the Create Custom Equation GUI. The GUI
contains two panes: a pane for creating linear equations and a pane for creating
general (nonlinear) equations. These panes are described below.

Linear Equations
Linear equations are defined as equations that are linear in the parameters.
For example, the polynomial library equations are linear. The Linear
Equations pane is shown below followed by a description of its parameters.

• Independent variable — Symbol representing the independent (predictor)

variable. The default symbol is x.
• Equation — Symbol representing the dependent (response) variable
followed by the linear equation. The default symbol is y.
- Unknown Coefficients — The unknown coefficients to be determined by
the fit. The default symbols are a, b, c, and so on.
- Terms — Functions that depend only on the independent variable and
constants. Note that if you attempt to define a term that contains a
coefficient to be fitted, an error is returned.
- Unknown constant coefficient — If selected, a constant term is included
in the equations to be fit. Otherwise, a constant term is not included.
- Add a term — Add a term to the equation. An unknown coefficient is
automatically added for each new term.
- Remove last term — Remove the last term added to the equation.

3-21
3 Fitting Data

• Equation — The custom equation.

• Equation name — The name of the equation. By default, the name is
automatically updated to be identical to the custom equation given by
Equation. If you override the default, the name is no longer automatically
updated.

General Equations
General (nonlinear) equations are defined as equations that are nonlinear in
the parameters, or are a combination of linear and nonlinear in the
parameters. For example, the exponential library equations are nonlinear. The
General Equations pane is shown below followed by a brief description of its
parameters.

• Independent variable — Symbol representing the independent (predictor)

variable. The default symbol is x.
• Equation — Symbol representing the dependent (response) variable
followed by the general equation. As you type in the terms of the equation,
the unknown coefficients, associated starting values, and constraints
automatically populate the table. By default, the starting values are
randomly selected on the interval [0,1] and are unconstrained.
You can immediately change the default starting values and constraints in
this table, or you can change them later using the Fit Options GUI.

3-22
 Parametric Fitting

• Equation name — The name of the equation. By default, the name is

automatically updated to be identical to the custom equation given by
Equation. If you override the default, the name is no longer automatically
updated.

Note that even if you define a linear equation, a nonlinear fitting procedure is
used. Although this is allowed by the toolbox, it is an inefficient process and can
result in less than optimal fitted coefficients. Instead, you should use the
Linear Equations pane to define the equation.

Specifying Fit Options

You specify fit options with the Fit Options GUI. The fit options for the
single-term exponential are shown below. The coefficient starting values and
constraints are for the census data.

Fitting method and algorithm

Finite differencing parameters

Fit convergence criteria

Coefficient parameters

The available GUI options depend on whether you are fitting your data using
a linear model, a nonlinear model, or a nonparametric fit type. All the options
described below are available for nonlinear models. Method, Robust, and
coefficient constraints (Lower and Upper) are available for linear models.
Interpolants and smoothing splines include Method, but no configurable
options.

3-23
3 Fitting Data

Fitting Method and Algorithm

• Method — The fitting method.
The method is automatically selected based on the library or custom model
you use. For linear models, the method is LinearLeastSquares. For
nonlinear models, the method is NonlinearLeastSquares.
• Robust — Specify whether to use the robust least squares fitting method.
The values are
- Off — Do not use robust fitting (default).
- On — Fit with default robust method (bisquare weights).
- LAR — Fit by minimizing the least absolute residuals (LAR).
- Bisquare — Fit by minimizing the summed square of the residuals, and
downweight outliers using bisquare weights. In most cases, this is the best
choice for robust fitting.
• Algorithm — Algorithm used for the fitting procedure:
- Trust-Region — This is the default algorithm and must be used if you
specify coefficient constraints.
- Levenberg-Marquardt — If the trust-region algorithm does not produce
a reasonable fit, and you do not have coefficient constraints, you should try
the Levenberg-Marquardt algorithm.
- Gauss-Newton — This algorithm is included for pedagogical reasons and
should be the last choice for most models and data sets.

Finite Differencing Parameters

• DiffMinChange — Minimum change in coefficients for finite difference
Jacobians. The default value is 10-8.
• DiffMaxChange — Maximum change in coefficients for finite difference
Jacobians. The default value is 0.1.

Fit Convergence Criteria

• MaxFunEvals — Maximum number of function (model) evaluations
allowed. The default value is 600.
• MaxIter — Maximum number of fit iterations allowed. The default value is
400.

3-24
 Parametric Fitting

• TolFun — Termination tolerance used on stopping conditions involving the

function (model) value. The default value is 10-6.
• TolX — Termination tolerance used on stopping conditions involving the
coefficients. The default value is 10-6.

Coefficient Parameters
• Unknowns — Symbols for the unknown coefficients to be fitted.
• StartPoint — The coefficient starting values. The default values depend on
the model. For rational, Weibull, and custom models, default values are
randomly selected within the range [0,1]. For all other nonlinear library
models, the starting values depend on the data set and are calculated
heuristically.
• Lower — Lower bounds on the fitted coefficients. The bounds are used only
with the trust region fitting algorithm. The default lower bounds for most
library models are -Inf, which indicates that the coefficients are
unconstrained. However, a few models have finite default lower bounds. For
example, Gaussians have the width parameter constrained so that it cannot
be less than 0.
• Upper — Upper bounds on the fitted coefficients. The bounds are used only
with the trust region fitting algorithm. The default upper bounds for all
library models are Inf, which indicates that the coefficients are
unconstrained.

For more information about these fit options, refer to “Optimization Options
Parameters” in the Optimization Toolbox documentation.

3-25
3 Fitting Data

Default Coefficient Parameters

The default coefficient starting points and constraints for library and custom
models are given below. If the starting points are optimized, then they are
calculated heuristically based on the current data set. Random starting points
are defined on the interval [0,1] and linear models do not require starting
points.
If a model does not have constraints, the coefficients have neither a lower
bound nor an upper bound. You can override the default starting points and
constraints by providing your own values using the Fit Options GUI.

Table 3-1: Default Starting Points and Constraints

Model Starting Points Constraints

Custom linear N/A None

Custom nonlinear Random None

Exponentials Optimized None

Fourier series Optimized None

Gaussians Optimized ci > 0

Polynomials N/A None

Power series Optimized None

Rationals Random None

Sum of sines Optimized bi > 0

Weibull Random a, b > 0

Note that the sum of sines and Fourier series models are particularly sensitive
to starting points, and the optimized values might be accurate for only a few
terms in the associated equations. For an example that overrides the default
starting values for the sum of sines model, refer to “Example: Sectioning
Periodic Data” on page 2-35.

3-26
 Parametric Fitting

Evaluating the Goodness of Fit

After fitting data with one or more models, you should evaluate the goodness
of fit. A visual examination of the fitted curve displayed in the Curve Fitting
Tool should be your first step. Beyond that, the toolbox provides these goodness
of fit measures for both linear and nonlinear parametric fits:

• Residuals
• Goodness of fit statistics
• Confidence and prediction bounds

You can group these measures into two types: graphical and numerical. The
residuals and prediction bounds are graphical measures, while the goodness of
fit statistics and confidence bounds are numerical measures.
Generally speaking, graphical measures are more beneficial than numerical
measures because they allow you to view the entire data set at once, and they
can easily display a wide range of relationships between the model and the
data. The numerical measures are more narrowly focused on a particular
aspect of the data and often try to compress that information into a single
number. In practice, depending on your data and analysis requirements, you
might need to use both types to determine the best fit.
Note that it is possible that none of your fits can be considered the best one. In
this case, it might be that you need to select a different model. Conversely, it is
also possible that all the goodness of fit measures indicate that a particular fit
is the best one. However, if your goal is to extract fitted coefficients that have
physical meaning, but your model does not reflect the physics of the data, the
resulting coefficients are useless. In this case, understanding what your data
represents and how it was measured is just as important as evaluating the
goodness of fit.

Residuals
The residuals from a fitted model are defined as the differences between the
response data and the fit to the response data at each predictor value.
residual = data - fit
You display the residuals in the Curve Fitting Tool by selecting the menu item
View->Residuals.

3-27
3 Fitting Data

Mathematically, the residual for a specific predictor value is the difference

between the response value y and the predicted response value ŷ .

r = y – ŷ
Assuming the model you fit to the data is correct, the residuals approximate
the random errors. Therefore, if the residuals appear to behave randomly, it
suggests that the model fits the data well. However, if the residuals display a
systematic pattern, it is a clear sign that the model fits the data poorly.
A graphical display of the residuals for a first degree polynomial fit is shown
below. The top plot shows that the residuals are calculated as the vertical
distance from the data point to the fitted curve. The bottom plot shows that the
residuals are displayed relative to the fit, which is the zero line.

12 Data
Linear Fit
10

0
0 1 2 3 4 5 6 7 8 9 10 11

Residuals
3
2
1
0
−1
−2
−3

0 1 2 3 4 5 6 7 8 9 10 11

The residuals appear randomly scattered around zero indicating that the
model describes the data well.

3-28
 Parametric Fitting

A graphical display of the residuals for a second-degree polynomial fit is shown

below. The model includes only the quadratic term, and does not include a
linear or constant term.

12 Data
Quadratic Fit
10

0
0 1 2 3 4 5 6 7 8 9 10 11

Residuals
3
2
1
0
−1
−2
−3

0 1 2 3 4 5 6 7 8 9 10 11

The residuals are systematically positive for much of the data range indicating
that this model is a poor fit for the data.

Goodness of Fit Statistics

After using graphical methods to evaluate the goodness of fit, you should
examine the goodness of fit statistics. The Curve Fitting Toolbox supports
these goodness of fit statistics for parametric models:

• The sum of squares due to error (SSE)

• R-square
• Adjusted R-square
• Root mean squared error (RMSE)

3-29
3 Fitting Data

For the current fit, these statistics are displayed in the Results list box in the
Fit Editor. For all fits in the current curve-fitting session, you can compare the
goodness of fit statistics in the Table of fits.

Sum of Squares Due to Error. This statistic measures the total deviation of the
response values from the fit to the response values. It is also called the summed
square of residuals and is usually labeled as SSE.
n
2
SSE = ∑ wi ( yi – ŷi )
i=1

A value closer to 0 indicates a better fit. Note that the SSE was previously
defined in “The Least Squares Fitting Method” on page 3-6.

R-Square. This statistic measures how successful the fit is in explaining the
variation of the data. Put another way, R-square is the square of the correlation
between the response values and the predicted response values. It is also called
the square of the multiple correlation coefficient and the coefficient of multiple
determination.
R-square is defined as the ratio of the sum of squares of the regression (SSR)
and the total sum of squares (SST). SSR is defined as
n
2
SSR = ∑ wi ( ŷi – y )
i=1

SST is also called the sum of squares about the mean, and is defined as
n
2
SST = ∑ wi ( yi – y )
i=1

where SST = SSR + SSE. Given these definitions, R-square is expressed as

SSR SSE
R-square = ------------- = 1 – -------------
SST SST
R-square can take on any value between 0 and 1, with a value closer to 1
indicating a better fit. For example, an R2 value of 0.8234 means that the fit
explains 82.34% of the total variation in the data about the average.

3-30
 Parametric Fitting

If you increase the number of fitted coefficients in your model, R-square might
increase although the fit may not improve. To avoid this situation, you should
use the degrees of freedom adjusted R-square statistic described below.
Note that it is possible to get a negative R-square for equations that do not
contain a constant term. If R-square is defined as the proportion of variance
explained by the fit, and if the fit is actually worse than just fitting a horizontal
line, then R-square is negative. In this case, R-square cannot be interpreted as
the square of a correlation.

Degrees of Freedom Adjusted R-Square. This statistic uses the R-square statistic
defined above, and adjusts it based on the residual degrees of freedom. The
residual degrees of freedom is defined as the number of response values n
minus the number of fitted coefficients m estimated from the response values.

v = n–m
v indicates the number of independent pieces of information involving the n
data points that are required to calculate the sum of squares. Note that if
parameters are bounded and one or more of the estimates are at their bounds,
then those estimates are regarded as fixed. The degrees of freedom is increased
by the number of such parameters.
The adjusted R-square statistic is generally the best indicator of the fit quality
when you add additional coefficients to your model.

SSE ( n – 1 )
adjusted R-square = 1 – -------------------------------
SST ( v – 1 )
The adjusted R-square statistic can take on any value less than or equal to 1,
with a value closer to 1 indicating a better fit.

Root Mean Squared Error. This statistic is also known as the fit standard error
and the standard error of the regression

RMSE = s = MSE
where MSE is the mean square error or the residual mean square

SSE
MSE = -------------
v
A RMSE value closer to 0 indicates a better fit.

3-31
3 Fitting Data

Confidence and Prediction Bounds

With the Curve Fitting Toolbox, you can calculate confidence bounds for the
fitted coefficients, and prediction bounds for new observations or for the fitted
function. Additionally, for prediction bounds, you can calculate simultaneous
bounds, which take into account all predictor values, or you can calculate
nonsimultaneous bounds, which take into account only individual predictor
values. The confidence bounds are numerical, while the prediction bounds are
displayed graphically.
The available confidence and prediction bounds are summarized below.

Table 3-2: Types of Confidence and Prediction Bounds

Interval Type Description

Fitted coefficients Confidence bounds for the fitted coefficients

New observation Prediction bounds for a new observation (response

value)

New function Prediction bounds for a new function value

Note Prediction bounds are often described as confidence bounds because

you are calculating a confidence interval for a predicted response.

Confidence and prediction bounds define the lower and upper values of the
associated interval, and define the width of the interval. The width of the
interval indicates how uncertain you are about the fitted coefficients, the
predicted observation, or the predicted fit. For example, a very wide interval
for the fitted coefficients can indicate that you should use more data when
fitting before you can say anything very definite about the coefficients.
The bounds are defined with a level of certainty that you specify. The level of
certainty is often 95%, but it can be any value such as 90%, 99%, 99.9%, and so
on. For example, you might want to take a 5% chance of being incorrect about
predicting a new observation. Therefore, you would calculate a 95% prediction
interval. This interval indicates that you have a 95% chance that the new
observation is actually contained within the lower and upper prediction
bounds.

3-32
 Parametric Fitting

Calculating and Displaying Confidence Bounds. The confidence bounds for fitted
coefficients are given by

C = b±t S
where b are the coefficients produced by the fit, t is the inverse of Student’s T
cumulative distribution function, and S is a vector of the diagonal elements
from the covariance matrix of the coefficient estimates, (XTX)-1s2. X is the
design matrix, XT is the transpose of X, and s2 is the mean squared error.
Refer to the tinv function, included with the Statistics Toolbox, for a
description of t. Refer to “Linear Least Squares” on page 3-6 for more
information about X and XT.
The confidence bounds are displayed in the Results list box in the Fit Editor
using the following format.
p1 = 1.275 (1.113, 1.437)

The fitted value for the coefficient p1 is 1.275, the lower bound is 1.113, the
upper bound is 1.437, and the interval width is 0.324. By default, the
confidence level for the bounds is 95%. You can change this level to any value
with the View->Confidence Level menu item in the Curve Fitting Tool.
You can calculate confidence intervals at the command line with the confint
function.

Calculating and Displaying Prediction Bounds. As mentioned previously, you can

calculate prediction bounds for a new observation or for the fitted curve. In
both cases, the prediction is based on an existing fit to the data. Additionally,
the bounds can be simultaneous and measure the confidence for all predictor
values, or they can be nonsimultaneous and measure the confidence only for a
single predetermined predictor value. If you are predicting a new observation,
nonsimultaneous bounds measure the confidence that the new observation lies
within the interval given a single predictor value. Simultaneous bounds
measure the confidence that a new observation lies within the interval
regardless of the predictor value.

3-33
3 Fitting Data

The nonsimultaneous prediction bounds for a new observation at the predictor

value x are given by

2
P n, o = ŷ ± t s + xSx'

where s2 is the mean squared error, t is the inverse of Student’s T cumulative

distribution function, and S is the covariance matrix of the coefficient
estimates, (XTX)-1s2. Note that x is defined as a row vector of the Jacobian
evaluated at a specified predictor value.
The simultaneous prediction bounds for a new observation and for all predictor
values are given by

2
P s, o = ŷ ± f s + xSx'

where f is the inverse of the F cumulative distribution function. Refer to the

finv function, included with the Statistics Toolbox, for a description of f.

The nonsimultaneous prediction bounds for the function at a single predictor

value x are given by

P n, f = ŷ ± t xSx'

The simultaneous prediction bounds for the function and for all predictor
values are given by

P s, f = ŷ ± f xSx'

You can graphically display prediction bounds two ways: using the Curve
Fitting Tool or using the Analysis GUI. With the Curve Fitting Tool, you can
display nonsimultaneous prediction bounds for new observations with
View->Prediction Bounds. By default, the confidence level for the bounds is
95%. You can change this level to any value with View->Confidence Level.
With the Analysis GUI, you can display nonsimultaneous prediction bounds for
the function or for new observations.
You can display numerical prediction bounds of any type at the command line
with the predint function.

3-34
 Parametric Fitting

To understand the quantities associated with each type of prediction interval,

recall that the data, fit, and residuals (random errors) are related through the
formula
data = fit + residuals
Suppose you plan to take a new observation at the predictor value xn+1. Call
the new observation yn+1(xn+1) and the associated error en+1. Then yn+1(xn+1)
satisfies the equation

yn + 1 ( xn + 1 ) = f ( xn + 1 ) + en + 1

where f(xn+1) is the true but unknown function you want to estimate at xn+1.
The likely values for the new observation or for the estimated function are
provided by the nonsimultaneous prediction bounds.
If instead you want the likely value of the new observation to be associated
with any predictor value, the previous equation becomes

yn + 1 ( x ) = f ( x ) + e

The likely values for this new observation or for the estimated function are
provided by the simultaneous prediction bounds.
The types of prediction bounds are summarized below.

Table 3-3: Types of Prediction Bounds

Type of Bound Associated Equation

Observation Nonsimultaneous yn+1(xn+1)

Simultaneous yn+1(x), globally for any x

Function Nonsimultaneous f(xn+1)

Simultaneous f(x), simultaneously for all x

The nonsimultaneous and simultaneous prediction bounds for a new

observation and the fitted function are shown below. Each graph contains three
curves: the fit, the lower confidence bounds, and the upper confidence bounds.
The fit is a single-term exponential to generated data and the bounds reflect a
95% confidence level. Note that the intervals associated with a new observation

3-35
3 Fitting Data

are wider than the fitted function intervals because of the additional
uncertainty in predicting a new response value (the fit plus random errors).

Nonsimultaneous bounds for function Nonsimultaneous bounds for observation

2.5 2.5
data data
fitted curve 2 fitted curve
2 prediction bounds prediction bounds
1.5
1.5
1
y

y
1
0.5
0.5 0

0 −0.5
0 2 4 6 8 10 0 2 4 6 8 10
x x

Simultaneous bounds for function Simultaneous bounds for observation

2.5 2.5
data data
fitted curve 2 fitted curve
2 prediction bounds prediction bounds
1.5
1.5
1
y

1
0.5
0.5 0

0 −0.5
0 2 4 6 8 10 0 2 4 6 8 10
x x

3-36
 Parametric Fitting

Example: Evaluating the Goodness of Fit

This example fits several polynomial models to generated data and evaluates
the goodness of fit. The data is cubic and includes a range of missing values.
rand('state',0)
x = [1:0.1:3 9:0.1:10]';
c = [2.5 -0.5 1.3 -0.1];
y = c(1) + c(2)*x + c(3)*x.^2 + c(4)*x.^3 + (rand(size(x))-0.5);

After you import the data, fit it using a cubic polynomial and a fifth degree
polynomial. The data, fits, and residuals are shown below. You display the
residuals in the Curve Fitting Tool with the View->Residuals menu item.

Both fits appear to

model the data well.

The residuals for both

fits appear to be
randomly distributed.

Both models appear to fit the data well, and the residuals appear to be
randomly distributed around zero. Therefore, a graphical evaluation of the fits
does not reveal any obvious differences between the two equations.

3-37
3 Fitting Data

The numerical fit results are shown below.

The cubic fit coefficients are

accurately known.

The quintic fit coefficients

are not accurately known.

As expected, the fit results for poly3 are reasonable because the generated data
is cubic. The 95% confidence bounds on the fitted coefficients indicate that they
are acceptably accurate. However, the 95% confidence bounds for poly5
indicate that the fitted coefficients are not known accurately.
The goodness of fit statistics are shown below. By default, the adjusted
R-square and RMSE statistics are not displayed in the Table of Fits. To
display these statistics, open the Table Options GUI by clicking the Table
options button. The statistics do not reveal a substantial difference between
the two equations.

The statistics do not reveal a substantial

difference between the two equations.

Open the Table Options GUI and

select Adj R-sq and RMSE.

3-38
 Parametric Fitting

The 95% nonsimultaneous prediction bounds for new observations are shown
below. To display prediction bounds in the Curve Fitting Tool, select the
View->Prediction Bounds menu item. Alternatively, you can view prediction
bounds for the function or for new observations using the Analysis GUI.

The prediction bounds for poly3 indicate that new observations can be
predicted accurately throughout the entire data range. This is not the case for
poly5. It has wider prediction bounds in the area of the missing data,
apparently because the data does not contain enough information to estimate
the higher degree polynomial terms accurately. In other words, a fifth-degree
polynomial overfits the data. You can confirm this by using the Analysis GUI
to compute bounds for the functions themselves.
The 95% prediction bounds for poly5 are shown below. As you can see, the
uncertainty in estimating the function is large in the area of the missing data.

3-39
3 Fitting Data

Therefore, you would conclude that more data must be collected before you can
make accurate predictions using a fifth-degree polynomial.

In conclusion, you should examine all available goodness of fit measures before
deciding on the best fit. A graphical examination of the fit and residuals should
always be your initial approach. However, some fit characteristics are revealed
only through numerical fit results, statistics, and prediction bounds.

3-40
 Parametric Fitting

Example: Rational Fit

This example fits measured data using a rational model. The data describes the
coefficient of thermal expansion for copper as a function of temperature in
degrees Kelvin.
To get started, load the thermal expansion data from the file hahn1.mat, which
is provided with the toolbox.
load hahn1

The workspace now contains two new variables, temp and thermex:

• temp is a vector of temperatures in degrees Kelvin.

• thermex is a vector of thermal expansion coefficients for copper.

Import these two variables into the Curve Fitting Tool and name the data set
CuThermEx.

For this data set, you will find the rational equation that produces the best fit.
As described in “Library Models” on page 3-16, rational models are defined as
a ratio of polynomials
n n–1
p1 x + p2 x + … + pn + 1
y = ------------------------------------------------------------------------
m m–1
x + q1 x + … + qm

where n is the degree of the numerator polynomial and m is the degree of the
denominator polynomial. Note that the rational equations are not associated
with physical parameters of the data. Instead, they provide a simple and
flexible empirical model that you can use for interpolation and extrapolation.

3-41
3 Fitting Data

As you can see by examining the shape of the data, a reasonable initial choice
for the rational model is quadratic/quadratic. The Fitting GUI configured for
this equation is shown below.

Begin the fitting process with a

quadratic/quadratic rational fit.

3-42
 Parametric Fitting

The data, fit, and residuals are shown below.

The fit clearly misses

some of the data.

The residuals show a

strong pattern indicating
a better fit is possible.

The fit clearly misses the data for the smallest and largest predictor values.
Additionally, the residuals show a strong pattern throughout the entire data
set indicating that a better fit is possible.

3-43
3 Fitting Data

For the next fit, try a cubic/cubic equation. The data, fit, and residuals are
shown below.

The fit exhibits several

discontinuities around the
zeros of the denominator.

The numerical results shown below indicate that the fit did not converge.

The fit did not converge, which

indicates that the model might
be a poor choice for the data.

3-44
 Parametric Fitting

Although the message in the Results window indicates that you might improve
the fit if you increase the maximum number of iterations, a better choice at this
stage of the fitting process is to use a different rational equation because the
current fit contains several discontinuities. These discontinuities are due to the
function blowing up at predictor values that correspond to the zeros of the
denominator.
As the next try, fit the data using a cubic/quadratic equation. The data, fit, and
residuals are shown below.

The fit is well behaved

over the entire data range.

The residuals are

randomly scattered
about zero.

The fit is well behaved over the entire data range, and the residuals are
randomly scattered about zero. Therefore, you can confidently use this fit for
further analysis.

3-45
3 Fitting Data

Example: Fitting with Custom Equations

You can define your own equations with the Create Custom Equation GUI. You
open this GUI one of two ways:

• From the Curve Fitting Tool, select Tools->Custom Equation.

• From the Fitting GUI, select Custom Equations from the Type of fit list,
then click the New Equation button.

The Create Custom Equation GUI contains two panes: one for creating linear
custom equations and one for creating general (nonlinear) custom equations.
These panes are described in the following examples.

Linear Equation: Legendre Polynomial Fit

This example fits data using several custom linear equations. The data is
generated, and is based on the nuclear reaction 12C(e,e’α)8Be. The equations
use sums of Legendre polynomial terms.
Consider an experiment in which 124 MeV electrons are scattered from 12C
nuclei. In the subsequent reaction, alpha particles are emitted and produce the
residual nuclei 8Be. By analyzing the number of alpha particles emitted as a
function of angle, you can deduce certain information regarding the nuclear
dynamics of 12C. The reaction kinematics are shown below.

e is the incident electron.

e’ 12
C is the carbon target.
q is the momentum transferred to 8Be.
θe’ e’ is the scattered electron.
12
C α is the emitted alpha particle.
e θe’ is the electron scattering angle.
q θα is the alpha scattering angle.

θα

The data is collected by placing solid state detectors at values of θα ranging

from 10o to 240o in 10o increments.

3-46
 Parametric Fitting

It is sometimes useful to describe a variable expressed as a function of angle in

terms of Legendre polynomials
∞

y( x) = ∑ an Pn ( x )
n=0

where Pn(x) is a Legendre polynomial of degree n, x is cos(θα), and an are the

coefficients of the fit. Refer to MATLAB’s legendre function for information
about generating Legendre polynomials.
For the alpha-emission data, you can directly associate the coefficients with the
nuclear dynamics by invoking a theoretical model, which is described in [8].
Additionally, the theoretical model introduces constraints for the infinite sum
shown above. In particular, by considering the angular momentum of the
reaction, a fourth-degree Legendre polynomial using only even terms should
describe the data effectively.
You can generate Legendre polynomials with Rodrigues’ formula:

1  d n 2 n
P n ( x ) = -----------
- ------- ( x – 1 )
n  dx
2 n!
The Legendre polynomials up to fourth degree are given below.

Table 3-4: Legendre Polynomials up to Fourth Degree

n Pn(x)

0 1

1 x

2 (1/2)(3x2– 1)

3 (1/2)(5x3 – 3x)

4 (1/8)(35x4 – 30x2 + 3)

3-47
3 Fitting Data

The first step is to load the 12C alpha-emission data from the file
carbon12alpha.mat, which is provided with the toolbox.
load carbon12alpha

The workspace now contains two new variables, angle and counts:

• angle is a vector of angles (in radians) ranging from 10o to 240o in 10o
increments.
• counts is a vector of raw alpha particle counts that correspond to the
emission angles in angle.

Import these two variables into the Curve Fitting Toolbox and name the data
set C12Alpha.
The Fit Editor for a custom equation fit type is shown below.

Specify a meaningful fit name,

the data set, and the type of fit.

Open the Create Custom

Equations GUI.

3-48
 Parametric Fitting

Fit the data using a fourth-degree Legendre polynomial with only even terms:

1 1
y 1 ( x ) = a 0 + a 2  --- ( 3x – 1 ) + a 4  --- ( 35x – 30x + 3 )
2 4 2
 2  8

Because the Legendre polynomials depend only on the predictor variable and
constants, you use the Linear Equations pane on the Create Custom Equation
GUI. This pane is shown below for the model given by y1(x). Note that because
angle is given in radians, the argument of the Legendre terms is given by
cos(θα).

Create a custom linear equation

using even Legendre terms up
to fourth degree.

Specify a meaningful
equation name.

3-49
3 Fitting Data

The fit and residuals are shown below. The fit appears to follow the trend of the
data well, while the residuals appear to be randomly distributed and do not
exhibit any systematic behavior.

The numerical fit results are shown below. The 95% confidence bounds indicate
that the coefficients associated with P0(x) and P4(x) are known fairly
accurately, but that the P2(x) coefficient has a relatively large uncertainty.

The coefficients associated with P0(x) and

P4(x) are known accurately, but the P2(x)
coefficient has a larger uncertainty.

3-50
 Parametric Fitting

To confirm the theoretical argument that the alpha-emission data is best

described by a fourth-degree Legendre polynomial with only even terms, fit the
data using both even and odd terms:

1
y 2 ( x ) = y 1 ( x ) + a 1 x + a3  --- ( 5x – 3x )
3
 2

The Linear Equations pane of the Create Custom Equation GUI is shown below
for the model given by y2(x).

Create a custom linear equation

using even and odd Legendre
terms up to fourth degree.

Click Add a term to add the odd

Legendre terms.
Specify a meaningful
equation name.

The numerical results indicate that the odd Legendre terms do not contribute
significantly to the fit, and the even Legendre terms are essentially unchanged
from the previous fit. This confirms that the initial model choice is the best one.

The odd Legendre coefficients should not be

included in the fit because their values are
small and their confidence bounds are large.

3-51
3 Fitting Data

General Equation: Fourier Series Fit

This example fits the ENSO data using several custom nonlinear equations.
The ENSO data consists of monthly averaged atmospheric pressure differences
between Easter Island and Darwin, Australia. This difference drives the trade
winds in the southern hemisphere.
As shown in “Example: Smoothing Data” on page 2-21, the ENSO data is
clearly periodic, which suggests it can be described by a Fourier series
∞
x x
y ( x ) = a0 + ∑ ai cos  2π ---c-i + bi sin  2π c----i
i=1

where ai and bi are the amplitudes, and ci are the periods (cycles) of the data.
The question to be answered in this example is how many cycles exist? As a
first attempt, assume a 12 month cycle and fit the data using one sine term and
one cosine term.

x x
y 1 ( x ) = a 0 + a 1 cos  2π ----- + b 1 sin  2π -----
 c 1  c 1

If the fit does not describe the data well, add additional sine and cosine terms
with unique period coefficients until a good fit is obtained.
Because there is an unknown coefficient c1 included as part of the
trigonometric function arguments, the equation is nonlinear. Therefore, you
must specify the equation using the General Equations pane of the Create
Custom Equation GUI. This pane is shown below for the equation given by
y1(x).

Assume one 12 month cycle.

By default, the coefficients are

unbounded and have random
starting values between 0 and 1.

Specify a meaningful
equation name.

3-52
 Parametric Fitting

Note that the toolbox includes the Fourier series as a nonlinear library
equation. However, the library equation does not meet the needs of this
example because its terms are defined as fixed multiples of the fundamental
frequency w. Refer to “Fourier Series” on page 3-16 for more information.
The numerical results shown below indicate that the fit does not describe the
data well. In particular, the fitted value for c1 is unreasonably small. Because
the starting points are randomly selected, your initial fit results might differ
from the results shown here.

To assist the fitting procedure, constrain c1 to a value between 10 and 14. To

define constraints for unknown coefficients, use the Fit Options GUI, which
you open by clicking the Fit options button in the Fitting GUI.

Constrain the cycle to be

between 10 and 14 months.

3-53
3 Fitting Data

The fit, residuals, and numerical results are shown below.

The fit for one cycle.

The residuals indicate that at

least one more cycle exists.

The numerical results

indicate a 12 month cycle.

The fit appears to be reasonable for some of the data points but clearly does not
describe the entire data set very well. As predicted, the numerical results
indicate a cycle of approximately 12 months. However, the residuals show a
systematic periodic distribution indicating that there are additional cycles that
you should include in the fit equation. Therefore, as a second attempt, add an
additional sine and cosine term to y1(x)

x x
y 2 ( x ) = y 1 ( x ) + a 2 cos  2π ----- + b 2 sin  2π -----
 c 2  c 2

and constrain the upper and lower bounds of c2 to be roughly twice the bounds
used for c1.

3-54
 Parametric Fitting

The fit, residuals, and numerical results are shown below.

The fit for two cycles.

The residuals indicate that

one more cycle might exist.

The numerical results indicate

an additional 22 month cycle.

The fit appears to be reasonable for most of the data points. However, the
residuals indicate that you should include another cycle to the fit equation.
Therefore, as a third attempt, add an additional sine and cosine term to y2(x)

x x
y 3 ( x ) = y 2 ( x ) + a 3 cos  2π ----- + b 3 sin  2π -----
 c 3  c 3

and constrain the lower bound of c3 to be roughly three times the value of c1.

3-55
3 Fitting Data

The fit, residuals, and numerical results are shown below.

The fit for three cycles.

The residuals appear

fairly random for most
of the data set.

The numerical results indicate

12, 22, and 44 month cycles.

The fit is an improvement over the previous two fits, and appears to account
for most of the cycles present in the ENSO data set. The residuals appear
random for most of the data, although a pattern is still visible indicating that
additional cycles may be present, or you can improve the fitted amplitudes.
In conclusion, Fourier analysis of the data reveals three significant cycles. The
annual cycle is the strongest, but cycles with periods of approximately 44 and
22 months are also present. These cycles correspond to El Nino and the
Southern Oscillation (ENSO).

3-56
 Parametric Fitting

General Equation: Gaussian Fit with Exponential Background

This example fits two poorly resolved Gaussian peaks on a decaying
exponential background using a general (nonlinear) custom model. To get
started, load the data from the file gauss3.mat, which is provided with the
toolbox.
load gauss3

The workspace now contains two new variables, xpeak and ypeak:

• xpeak is a vector of predictor values.

• ypeak is a vector of response values.

Import these two variables into the Curve Fitting Toolbox and accept the
default data set name ypeak vs. xpeak.
You will fit the data with the following equation
x – b1 2 x – b2 2
–  -------------- –  --------------
– bx  c1   c2 
y ( x ) = ae + a1 e + a2 e

where ai are the peak amplitudes, bi are the peak centroids, and ci are related
to the peak widths. Because there are unknown coefficients included as part of
the exponential function arguments, the equation is nonlinear. Therefore, you
must specify the equation using the General Equations pane of the Create
Custom Equation GUI. This pane is shown below for y(x).

Two Gaussian peaks on an

exponential background.

By default, the coefficients are

unbounded and have random
starting values between 0 and 1.

3-57
3 Fitting Data

The data, fit, and numerical fit results are shown below. Clearly, the fit is poor.

Because the starting points are randomly selected, your initial fit results might
differ from the results shown here.

3-58
 Parametric Fitting

The results include this warning message.

Fit computation did not converge:
Maximum number of function evaluations exceeded. Increasing
MaxFunEvals (in fit options) may allow for a better fit, or
the current equation may not be a good model for the data.

To improve the fit for this example, specify reasonable starting points for the
coefficients. Deducing the starting points is particularly easy for the current
model because the Gaussian coefficients have a straightforward interpretation
and the exponential background is well defined. Additionally, as the peak
amplitudes and widths cannot be negative, constrain a1, a2, c1, and c2 to be
greater then zero.
To define starting values and constraints for unknown coefficients, use the Fit
Options GUI, which you open by clicking the Fit options button. The starting
values and constraints are shown below.

Specify reasonable coefficient

starting values and constraints.

3-59
3 Fitting Data

The data, fit, residuals, and numerical results are shown below.

3-60
 Parametric Fitting

Example: Robust Fit

This example fits data that is assumed to contain one outlier. The data consists
of the 2000 United States presidential election results for the state of Florida.
The fit model is a first degree polynomial and the fit method is robust linear
least squares with bisquare weights.
In the 2000 presidential election, many residents of Palm Beach County,
Florida, complained that the design of the election ballot was confusing, which
they claim led them to vote for the Reform candidate Pat Buchanan instead of
the Democratic candidate Al Gore. The so-called “butterfly ballot” was used
only in Palm Beach County and only for the election-day ballots for the
presidential race. As you will see, the number of Buchanan votes for Palm
Beach is far removed from the bulk of data, which suggests that the data point
should be treated as an outlier.
To get started, load the Florida election result data from the file flvote2k.mat,
which is provided with the toolbox.
load flvote2k

The workspace now contains these three new variables:

• buchanan is a vector of votes for the Reform Party candidate Pat Buchanan.
• bush is a vector of votes for the Republican Party candidate George Bush.
• gore is a vector of votes for the Democratic Party candidate Al Gore.

Each variable contains 68 elements, which correspond to the 67 Florida

counties plus the absentee ballots. The names of the counties are given in the
variable counties. From these variables, create two data sets with the
Buchanan votes as the response data: buchanan vs. bush and buchanan vs.
gore.

For this example, assume that the relationship between the response and
predictor data is linear with an offset of zero.
buchanan votes = (bush votes)(m1)
buchanan votes = (gore votes)(m2)
m1 is the number of Bush votes expected for each Buchanan vote, and m2 is
the number of Gore votes expected for each Buchanan vote.

3-61
3 Fitting Data

To create a first-degree polynomial equation with zero offset, you must create
a custom linear equation. As described in “Example: Fitting with Custom
Equations” on page 3-46, you can create a custom equation using the Fitting
GUI by selecting Custom Equations from the Type of fit list, and then clicking
the New Equation button.
The Linear Equations pane of the Create Custom Equation GUI is shown
below.

Create a first-degree
polynomial with zero offset.

Clear this check box.

Assign a meaningful name
to the equation.

Before fitting, you should exclude the data point associated with the absentee
ballots from each data set because these voters did not use the butterfly ballot.
As described in “Marking Outliers” on page 2-27, you can exclude individual
data points from a fit either graphically or numerically using the Exclude GUI.
For this example, you should exclude the data numerically. The index of the
absentee ballot data is given by
ind = find(strcmp(counties,'Absentee Ballots'))
ind =
68

3-62
 Parametric Fitting

The Exclude GUI is shown below.

Mark the absentee

votes to be excluded.

The exclusion rule is named AbsenteeVotes. You use the Fitting GUI to
associate an exclusion rule with the data set to be fit.
For each data set, perform a robust fit with bisquare weights using the
FlaElection equation defined above. For comparison purposes, also perform a
regular linear least squares fit. Refer to “Robust Least Squares” on page 3-11
for a description of the robust fitting methods provided by the toolbox.
You can identify the Palm Beach County data in the scatter plot by using the
data tips feature, and knowing the index number of the data point.
ind = find(strcmp(counties,'Palm Beach'))
ind =
50

3-63
3 Fitting Data

The Fit Editor and the Fit Options GUI are shown below for a robust fit.

Associate the excluded

absentee votes with the fit.

Choose robust fitting

with bisquare weights.

Open the Fit Options GUI.

The data, robust and regular least squares fits, and residuals for the buchanan
vs. bush data set are shown below.

The data tip shows that

Buchanan received 3411
votes in Palm Beach County.

The Palm Beach County

residual is very large.

The Miami/Dade County

residual is also very large.

3-64
 Parametric Fitting

The graphical results show that the linear model is reasonable for the majority
of data points, and the residuals appear to be randomly scattered around zero.
However, two residuals stand out. The largest residual corresponds to Palm
Beach County. The other residual is at the largest predictor value, and
corresponds to Miami/Dade County.
The numerical results are shown below. The inverse slope of the robust fit
indicates that Buchanan should receive one vote for every 197.4 Bush votes.

The data, robust and regular least squares fits, and residuals for the buchanan
vs. gore data set are shown below.

The Palm Beach County

residual is very large.

The Miami/Dade and

Broward County residuals
are also very large.

3-65
3 Fitting Data

Again, the graphical results show that the linear model is reasonable for the
majority of data points, and the residuals appear to be randomly scattered
around zero. However, three residuals stand out. The largest residual
corresponds to Palm Beach County. The other residuals are at the two largest
predictor values, and correspond to Miami/Dade County and Broward County.
The numerical results are shown below. The inverse slope of the robust fit
indicates that Buchanan should receive one vote for every 189.3 Gore votes.

Using the fitted slope value, you can determine the expected number of votes
that Buchanan should have received for each fit. For the Buchanan versus
Bush data, you evaluate the fit at a predictor value of 152,951. For the
Buchanan versus Gore data, you evaluate the fit at a predictor value of
269,732. These results are shown below for both data sets and both fits.

Table 3-5: Expected Buchanan Votes in Palm Beach County

Data Set Fit Expected Buchanan Votes

Buchanan vs. Bush Regular least squares 814

Robust least squares 775

Buchanan vs. Gore Regular least squares 1246

Robust least squares 1425

The robust results for the Buchanan versus Bush data suggest that Buchanan
received 3411 – 775 = 2636 excess votes, while robust results for the Buchanan
versus Gore data suggest that Buchanan received 3411 – 1425 = 1986 excess
votes.

3-66
 Parametric Fitting

The margin of victory for George Bush is given by

margin = sum(bush)–sum(gore)
margin =

537

Therefore, the voter intention comes into play because in both cases, the
margin of victory is less than the excess Buchanan votes.
In conclusion, the analysis of the 2000 United States presidential election
results for the state of Florida suggests that the Reform Party candidate
received an excess number of votes in Palm Beach County, and that this excess
number was a crucial factor in determining the election outcome. However,
additional analysis is required before a final conclusion can be made.

3-67
3 Fitting Data

Nonparametric Fitting
In some cases, you are not concerned about extracting or interpreting fitted
parameters. Instead, you might simply want to draw a smooth curve through
your data. Fitting of this type is called nonparametric fitting. The Curve Fitting
Toolbox supports these nonparametric fitting methods:

• Interpolants — Estimate values that lie between known data points.

• Smoothing spline — Create a smooth curve through the data. You adjust the
level of smoothness by varying a parameter that changes the curve from a
least squares straight-line approximation to a cubic spline interpolant.

For more information about interpolation, refer to “Polynomials and

Interpolation” and the interp1 function in the MATLAB documentation. For
more information about smoothing splines, refer to “Tutorial” and the csaps
function in the Spline Toolbox documentation.

Interpolants
Interpolation is a process for estimating values that lie between known data
points. The supported interpolant methods are shown below.

Table 3-6: Interpolant Methods

Method Description

Linear Linear interpolation. This method fits a different

linear polynomial between each pair of data points.

Nearest neighbor Nearest neighbor interpolation. This method sets the

value of an interpolated point to the value of the
nearest data point. Therefore, this method does not
generate any new data points.

Cubic spline Cubic spline interpolation. This method fits a different

cubic polynomial between each pair of data points.

Shape-preserving Piecewise cubic Hermite interpolation (PCHIP). This

method preserves monotonicity and the shape of the
data.

3-68
 Nonparametric Fitting

The type of interpolant you should use depends on the characteristics of the
data being fit, the required smoothness of the curve, speed considerations,
postfit analysis requirements, and so on. The linear and nearest neighbor
methods are fast, but the resulting curves are not very smooth. The cubic spline
and shape-preserving methods are slower, but the resulting curves are often
very smooth.
For example, the nuclear reaction data from the file carbon12alpha.mat is
shown below with a nearest neighbor interpolant fit and a shape-preserving
(PCHIP) interpolant fit. Clearly, the nearest neighbor interpolant does not
follow the data as well as the shape-preserving interpolant. The difference
between these two fits can be important if you are interpolating. However, if
you want to integrate the data to get a sense of the total unormalized strength
of the reaction, then both fits provide nearly identical answers for reasonable
integration bin widths.

350
C12Alpha
nearest
pchip
300

250

200
counts

150

100

50

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
angle

3-69
3 Fitting Data

Note Goodness of fit statistics, prediction bounds, and weights are not
defined for interpolants. Additionally, the fit residuals are always zero (within
computer precision) because interpolants pass through the data points.

Interpolants are defined as piecewise polynomials because the fitted curve is

constructed from many “pieces.” For cubic spline and PCHIP interpolation,
each piece is described by four coefficients, which are calculated using a cubic
(third-degree) polynomial. Refer to the spline function for more information
about cubic spline interpolation. Refer to the pchip function for more
information about shape-preserving interpolation, and for a comparison of the
two methods.
Parametric polynomial fits result in a global fit where one set of fitted
coefficients describes the entire data set. As a result, the fit can be erratic.
Because piecewise polynomials always produce a smooth fit, they are more
flexible than parametric polynomials and can be effectively used for a wider
range of data sets.

Smoothing Spline
If your data is noisy, you might want to fit it using a smoothing spline.
Alternatively, you can use one of the smoothing methods described in
“Smoothing Data” on page 2-9.
The smoothing spline s is constructed for the specified smoothing parameter p
and the specified weights wi. The smoothing spline minimizes
2 2
2 d s 
p ∑ ∫
w i ( y i – s ( x i ) ) + ( 1 – p )  2  dx
dx 
i

If the weights are not specified, they are assumed to be 1 for all data points.
p is defined between 0 and 1. p = 0 produces a least squares straight line fit to
the data, while p = 1 produces a cubic spline interpolant. If you do not specify
the smoothing parameter, it is automatically selected in the “interesting
range.” The interesting range of p is often near 1/(1+h3/6) where h is the
average spacing of the data points, and it is typically much smaller than the
allowed range of the parameter. Because smoothing splines have an associated

3-70
 Nonparametric Fitting

parameter, you can consider these fits to be parametric. However, smoothing

splines are also piecewise polynomials like cubic spline or shape-preserving
interpolants and are considered a nonparametric fit type in this guide.

Note The smoothing spline algorithm used by the Curve Fitting Toolbox is
based on the csaps function included with the Spline Toolbox. Refer to the
csaps reference pages for detailed information about smoothing splines.

The nuclear reaction data from the file carbon12alpha.mat is shown below
with three smoothing spline fits. The default smoothing parameter (p = 0.99)
produces the smoothest curve. The cubic spline curve (p = 1) goes through all
the data points, but is not quite as smooth. The third curve (p = 0.95) misses
the data by wide margin and illustrates how small the “interesting range” of p
can be.

350
C12Alpha
p=default
p=1
300 p=0.95

250

200
counts

150

100

50

−50
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
angle

3-71
3 Fitting Data

Example: Nonparametric Fit

This example fits the following data using a cubic spline interpolant and
several smoothing splines.
rand('state',0);
x = (4*pi)*[0 1 rand(1,25)];
y = sin(x) + .2*(rand(size(x))-.5);

As shown below, you can fit the data with a cubic spline by selecting
Interpolant from the Type of fit list.

The results shown below indicate that goodness of fit statistics are not defined
for interpolants.

As described in “Interpolants” on page 3-68, cubic spline interpolation is

defined as a piecewise polynomial that results in a structure of coefficients. The
number of “pieces” in the structure is one less than the number of fitted data
points, and the number of coefficients for each piece is four because the
polynomial degree is three. The toolbox does not allow you to access the
structure of coefficients.

3-72
 Nonparametric Fitting

As shown below, you can fit the data with a smoothing spline by selecting
Smoothing Spline in the Type of fit list.

The default smoothing

parameter is based on
the data set you fit.

The level of smoothness is given by the Smoothing Parameter. The default

smoothing parameter value depends on the data set, and is automatically
calculated by the toolbox after you click the Apply button.
For this data set, the default smoothing parameter is close to 1, indicating that
the smoothing spline is nearly cubic and comes very close to passing through
each data point. Create a fit for the default smoothing parameter and name it
Smooth1. If you do not like the level of smoothing produced by the default
smoothing parameter, you can specify any value between 0 and 1. A value of 0
produces a piecewise linear polynomial fit, while a value of 1 produces a
piecewise cubic polynomial fit, which passes through all the data points. For
comparison purposes, create another smoothing spline fit using a smoothing
parameter of 0.5 and name the fit Smooth2.
The numerical results for the smoothing spline fit Smooth1 are shown below.

3-73
3 Fitting Data

The data and fits are shown below. The default abscissa scale was increased to
show the fit behavior beyond the data limits. You change the axes limits with
Tools->Axes Limit Control menu item.

The cubic spline and default

smoothing spline results are
similar for interior points.

The cubic spline and default

smoothing spline results
diverge at the end points.

The default smoothing

parameter produces the
smoothest result.

Note that the default smoothing parameter produces the smoothest curve. As
the smoothing parameter increases beyond the default value, the associated
curve approaches the cubic spline curve.

3-74
 Selected Bibliography

Selected Bibliography
[1] Draper, N.R and H. Smith, Applied Regression Analysis, 3rd Ed., John
Wiley & Sons, New York, 1998.
[2] Bevington, P.R. and D.K. Robinson, Data Reduction and Error Analysis for
the Physical Sciences, 2nd Ed., WCB/McGraw-Hill, Boston, 1992.
[3] Daniel, C. and F.S. Wood, Fitting Equations to Data, John Wiley & Sons,
New York, 1980.
[4] Branch, M.A., T.F. Coleman, and Y. Li, “A Subspace, Interior, and
Conjugate Gradient Method for Large-Scale Bound-Constrained Minimization
Problems,” SIAM Journal on Scientific Computing, Vol. 21, Number 1, pp.
1-23, 1999.
[5] Levenberg, K., “A Method for the Solution of Certain Problems in Least
Squares,” Quart. Appl. Math, Vol. 2, pp. 164-168, 1944.
[6] Marquardt, D., “An Algorithm for Least Squares Estimation of Nonlinear
Parameters,” SIAM J. Appl. Math, Vol. 11, pp. 431-441, 1963.
[7] DuMouchel, W. and F. O’Brien, “Integrating a Robust Option into a
Multiple Regression Computing Environment,” in Computing Science and
Statistics: Proceedings of the 21st Symposium on the Interface, (K. Berk and L.
Malone, eds.), American Statistical Association, Alexandria, VA, pp. 297-301,
1989.
[8] DeAngelis, D.J., J.R. Calarco, J.E. Wise, H.J. Emrich, R. Neuhausen, and
H. Weyand, “Multipole Strength in 12C from the (e,e'α) Reaction for
Momentum Transfers up to 0.61 fm-1,” Phys. Rev. C, Vol. 52, Number 1, pp.
61-75 (1995).

3-75
3 Fitting Data

3-76
 4

Function Reference

This chapter describes the toolbox M-file functions that you use directly. A number of other M-file
helper functions are provided with this toolbox to support the functions listed below. These helper
functions are not documented because they are not intended for direct use.

Functions – By Category Contains a series of tables that group functions by category

(p. 4-78)
Functions – Alphabetical Lists all the functions alphabetically
List (p. 4-80)
4 Function Reference

Functions – By Category
Fitting Data
cfit Create a cfit object
fit Fit data using a library or custom model, a smoothing spline,
or an interpolant
fitoptions Create or modify a fit options object
fittype Create a fit type object

Getting Information and Help

cflibhelp Display information about library models, splines, and
interpolants
disp Display descriptive information for Curve Fitting Toolbox
objects

Getting and Setting Properties

get Return properties for a fit options object
set Configure or display property values for a fit options object

Preprocessing Data
excludedata Specify data to be excluded from a fit
smooth Smooth the response data

4-78
 Functions – By Category

Postprocessing Data
confint Compute confidence bounds for fitted coefficients
differentiate Differentiate a fit result object
integrate Integrate a fit result object
predint Compute prediction bounds for new observations or for the
function

General Purpose
cftool Open the Curve Fitting Tool
datastats Return descriptive statistics about the data
feval Evaluate a fit result object or a fit type object
plot Plot data, fit, prediction bounds, outliers, and residuals

4-79
4 Function Reference

Functions – Alphabetical List

This section contains detailed descriptions of all toolbox functions. Each
function reference page contains some or all of this information:

• The function name

• The purpose of the function
• The function syntax
All valid input argument and output argument combinations are shown. In
some cases, an ellipsis (. . .) is used for the input arguments. This means that
all preceding input argument combinations are valid for the specified output
argument(s).
• A description of each argument
• A description of the function
• Additional remarks about usage
• An example of usage
• Related functions

4-80
 cfit

Purpose 4cfit
Create a cfit object

Syntax fmodel = cfit(ftype,coef1,coef2,...)

Arguments ftype A fit type object representing a custom or library model.

coef1,coef2,... The model coefficients.

fmodel The cfit object.

Description fmodel = cfit(ftype,coef1,coef2,...) creates the cfit object fmodel based

on the custom or library model specified by ftype, and with the coefficients
specified by coef1, coef2, and so on. You create ftype with the fittype
function.

Remarks cfit is called by the fit function. You should call cfit directly if you want to
assign coefficients and problem parameters to a model without performing a
fit.

Example Create a fit type object and assign values to the coefficients and to the problem
parameter.
m = fittype('a*x^2+b*exp(n*x)','prob','n');
f = cfit(m,pi,10.3,3);

See Also fit, fittype

4-81
cflibhelp

Purpose 4cflibhelp
Display information about library models, splines, and interpolants

Syntax cflibhelp
cflibhelp group

Arguments group The name of the fit type group.

Description cflibhelp displays the names, equations, and descriptions for all the fit types
in the curve fitting library. You can use the fit type name as an input parameter
to the fit, cfit, and fittype functions.

cflibhelp group displays the names, equations, and descriptions for the fit
type group specified by group. The supported fit type groups are given below.

Group Description

distribution Distribution models such as Weibull

exponential One-term and two-term exponential equations

fourier Sums of sine and cosine equations up to eight terms

gaussian Sums of Gaussian equations up to eight terms

interpolant Interpolant fit types including linear, nearest neighbor,

cubic spline, and shape-preserving interpolation

polynomial Polynomial equations up to ninth degree

power One-term and two-term power equations

rational Ratios of polynomial equations up to degree 5 in both

numerator and denominator

sin Sums of sine equations up to eight terms

spline Cubic spline and smoothing spline fit types

For more information about the toolbox library models, refer to “Library
Models” on page 3-16. For more information about the toolbox library
interpolants and splines, refer to “Nonparametric Fitting” on page 3-68.

4-82
 cflibhelp

Example Display the names and descriptions for the spline fit type group.
cflibhelp spline

SPLINES

SPLINETYPE DESCRIPTION

cubicspline cubic interpolating spline

smoothingspline smoothing spline

Display the model names and equations for the polynomial fit type group.
cflibhelp polynomial

POLYNOMIAL MODELS

MODELNAME EQUATION

poly1 Y = p1*x+p2
poly2 Y = p1*x^2+p2*x+p3
poly3 Y = p1*x^3+p2*x^2+...+p4
...
poly9 Y = p1*x^9+p2*x^8+...+p10

See Also cfit, fit, fittype

4-83
cftool

Purpose 4cftool
Open the Curve Fitting Tool

Syntax cftool
cftool(xdata,ydata)

Arguments xdata A vector of predictor data.

ydata A vector of response data.

Description cftool opens the Curve Fitting Tool.

cftool(xdata,ydata) opens the Curve Fitting Tool with predictor data

specified by xdata and response data specified by ydata. xdata and ydata must
be vectors of the same size. Infs and NaNs are ignored because you cannot fit
data containing these values. Additionally, only the real component of a
complex value is used.

Remarks The Curve Fitting Tool is a graphical user interface (GUI) that allows you to

• Visually explore data and fits as scatter plots

• Graphically evaluate the goodness of fit using residuals and prediction
bounds
• Access GUIs for importing, preprocessing, and fitting data, and for plotting
and analyzing fits to the data

4-84
 cftool

The Curve Fitting Tool is shown below. The data is from the census MAT-file,
and the fit is a quadratic polynomial. The residuals are shown as a line plot
below the data and fit.

The Curve Fitting Tool provides several features that facilitate data and fit
exploration. Refer to “Viewing Data” on page 2-6 for a description of these
features.
By clicking the Data, Fitting, Exclude, Plotting, or Analysis buttons, you can
open the associated GUIs, which are described below. For a complete example
that uses many of these GUIs, refer to Chapter 1, “Getting Started with the
Curve Fitting Toolbox.”

4-85
cftool

The Data GUI

The Data GUI allows you to

• Import, preview, name, and delete data sets

• Smooth noisy data

The Data GUI is shown below with the census data loaded.

Refer to Chapter 2, “Importing, Viewing, and Preprocessing Data” for more

information about the Data GUI.

4-86
 cftool

The Fitting GUI

The Fitting GUI allows you to

• Fit data using a parametric or nonparametric equation

• Examine and compare fit results including fitted coefficient values and
goodness of fit statistics
• Keep track of all the data sets and fits for the current session

The Fitting GUI shown below displays the results of fitting the census data to
a quadratic polynomial.

4-87
cftool

The Exclude GUI

The Exclude GUI allows you to create exclusion rules for a data set. An
exclusion rule identifies data to be excluded while fitting. The excluded data
can be individual data points, or a section of predictor or response data. The
Exclude GUI shown below indicates that the first two data points of the census
data are marked for exclusion, and that this exclusion rule is named exc1.

The Plotting GUI

The Plotting GUI allows you to control the data sets and fits displayed by the
Curve Fitting Tool. The Plotting GUI shown below indicates that the census
data and the fit poly2 are displayed by the Curve Fitting Tool.

4-88
 cftool

The Analysis GUI

The Analysis GUI allows you to

• Evaluate (interpolate or extrapolate), differentiate, or integrate a fit

• Plot the analysis results and the data set

The Analysis GUI shown below displays the numerical results of extrapolating
the census data from the year 2000 to the year 2050 in 10-year increments.

Refer to “Analyzing the Fit” on page 1-15 for an example that uses the Analysis
GUI.

4-89
confint

Purpose 4confint
Compute confidence bounds for fitted coefficients

Syntax ci = confint(fresult)
ci = confint(fresult,level)

Arguments fresult A fit result object.

level The confidence level.

ci An array of confidence bounds.

Description ci = confint(fresult) returns 95% confidence bounds to ci for the fit

coefficients associated with fresult. fresult is the fit result object returned
by the fit function. ci is a 2-by-n array where n is the number of coefficients
associated with fresult. The top row of the array contains the lower bound,
while the bottom row of the array contains the upper bound for each coefficient

ci = confint(fresult,level) returns the confidence level specified by level

to ci. You specify level on the interval (0,1). For example, if level is 0.99, then
99% confidence bounds are calculated.

Remarks To calculate confidence bounds, confint uses R-1 (the inverse R factor from QR
decomposition of the Jacobian), the degrees of freedom for error, and the root
mean squared error. This information is automatically returned by the fit
function and contained within the fit result object.
The fit function automatically returns 95% confidence bounds for coefficients.
If coefficients are bounded and one or more of the estimates are at their bounds,
those estimates are regarded as fixed and do not have confidence bounds. Note
that you cannot calculate confidence bounds for the smoothing spline and
interpolant fit types.

4-90
 confint

Example Fit the census data to a second-degree polynomial. The display for fresult
includes the 95% confidence bounds for the fitted coefficients.
load census
fresult = fit(cdate,pop,'poly2')

fresult =
Linear model Poly2:
fresult(x) = p1*x^2 + p2*x + p3
Coefficients (with 95% confidence bounds):
p1 = 0.006541 (0.006124, 0.006958)
p2 = -23.51 (-25.09, -21.93)
p3 = 2.113e+004 (1.964e+004, 2.262e+004)

Calculate 95% confidence bounds for the fitted coefficients using confint.
ci = confint(fresult,0.95)
ci =

0.0061242 -25.086 19641

0.0069581 -21.934 22618

Note that the fit display and the array returned by confint present the
confidence bounds using slightly different formats. The fit display mimics an
n-by-3 array where n is the number of coefficients, the first column is the
coefficient variable, the second column is the fitted coefficient value, and the
third column is the lower and upper bound. confint returns a 2-by-n array
where the top row contains the lower bound and the bottom row contains the
upper bound for each coefficient.

See Also fit

4-91
datastats

Purpose 4datastats
Return descriptive statistics about the data

Syntax xds = datastats(xdata)

[xds,yds] = datastats(xdata,ydata)

Arguments xdata A column vector of predictor data.

ydata A column vector of response data.

xds A structure containing descriptive statistics for xdata.

yds A structure containing descriptive statistics for ydata.

Description xds = datastats(xdata) returns statistics for xdata to the structure xds. The
structure contains the fields shown below.

Field Description

num The number of data values

max The maximum data value

min The minimum data value

mean The mean value of the data

median The median value of the data

range The range of the data

std The standard deviation of the data

[xds,yds] = datastats(xdata,ydata) returns statistics for xdata and ydata

to the structures xds and yds, respectively. xds and yds contain the fields
shown above. xdata and ydata are column vectors of the same size.

Remarks If xdata or ydata contains complex values, only the real part of the value is
used in the statistics computations. If the data contains Infs or NaNs, they are
processed using the usual MATLAB rules.

4-92
 datastats

Example Return data statistics for the census data.

load census
[xds,yds] = datastats(cdate,pop)

xds =

num: 21
max: 1990
min: 1790
mean: 1890
median: 1890
range: 200
std: 62.048

yds =

num: 21
max: 248.7
min: 3.9
mean: 85.729
median: 62.9
range: 244.8
std: 78.601

4-93
differentiate

Purpose 4differentiate
Differentiate a fit result object

Syntax deriv1 = differentiate(fitresult,x)

[deriv1,deriv2] = differentiate(...)

Arguments fresult A fit result object.

x A column vector of values at which fresult is differentiated.

deriv1 A column vector of first derivatives.

deriv2 A column vector of second derivatives.

Description deriv1 = differentiate(fitresult,x) differentiates the fit result object

fresult at the points specified by x and returns the result to deriv1. You can
generate fresult with the fit function or the cfit function.

[deriv1,deriv2] = differentiate(...) computes the first derivative

deriv1, and the second derivative deriv2 for the specified fit result object.

Remarks For library equations with closed forms, analytic derivatives are calculated.
For all other equations, the first derivative is calculated using the central
difference quotient

yx + h – yx – h
y’ = ---------------------------------
2h
where x is the predictor value at which the derivative is calculated, h is a small
number, yx+h is fresult evaluated at x+h, and yx-h is fresult evaluated at x-h.
The second derivative is calculated using the expression

y x + h + y x – h – 2y x
y’’ = ------------------------------------------------
2
h

4-94
 differentiate

Example Create a noisy sine wave on the interval [0, 4π].

rand('state',0);
x = linspace(0,4*pi,200)';
y = sin(x) + (rand(size(x))-0.5)*0.2;

Create a custom fit type, and fit the data using reasonable starting values.
ftype = fittype('a*sin(b*x)');
fopts = fitoptions('Method','Nonlinear','start',[1 1]);
fit1 = fit(x,y,ftype,fopts);

Calculate the first derivative for each value of x.

deriv1 = differentiate(fit1,x);

Plot the data, the fit to the data, and the first derivatives.
plot(fit1,'k-',x,y,'b.');hold on
plot(x,deriv1,'ro')
legend('data','fitted curve','derivatives')

1.5
data
fitted curve
derivatives

0.5

0
y

−0.5

−1

−1.5
0 2 4 6 8 10 12 14
x

See Also cfit, fit, integrate

4-95
disp

Purpose 4disp
Display descriptive information for Curve Fitting Toolbox objects

Syntax obj
disp(obj)

Arguments obj A Curve Fitting Toolbox object.

Description obj or disp(obj) displays descriptive information for obj. You can create obj
with the fit or cfit function, the fitoptions function, or the fittype
function.

Example The display for a custom fit type object is shown below.
ftype = fittype('a*x^2+b*x+c+d*exp(-e*x)')

ftype =
General model:
ftype(a,b,c,d,e,x) = a*x^2+b*x+c+d*exp(-e*x)

The display for a fit options object is shown below.

fopts = fitoptions('Method','Nonlinear','Normalize','on')

fopts =
Normalize: 'on'
Exclude: []
Weights: []
Method: 'NonlinearLeastSquares'
Robust: 'Off'
StartPoint: []
Lower: []
Upper: []
Algorithm: 'Trust-Region'
DiffMinChange: 1e-008
DiffMaxChange: 0.1
Display: 'Notify'
MaxFunEvals: 600
MaxIter: 400
TolFun: 1e-006
TolX: 1e-006

4-96
 disp

Note that all fit types have the Normalize, Exclude, Weights, and Method fit
options. Additional fit options are available depending on the Method value. For
example, if Method is SmoothingSpline, the SmoothingParam fit option is
available.
The display for a fit result object is shown below.
fresult = fit(cdate,pop,ftype,fopts)

Warning: Start point not provided, choosing random start point.

Maximum number of function evaluations exceeded. Increasing
MaxFunEvals (in fit options) may allow for a better fit, or
the current equation may not be a good model for the data.

fresult =

General model:
fresult(x) = a*x^2+b*x+c+d*exp(-e*x)
where x is normalized by mean 1890 and std 62.05
Coefficients (with 95% confidence bounds):
a = 21.14 (-27.61, 69.89)
b = 64.49 (-188.5, 317.4)
c = 49.92 (-421.5, 521.4)
d = 11.96 (-458, 481.9)
e = -0.7745 (-10.25, 8.701)

See Also cfit, fit, fitoptions, fittype

4-97
excludedata

Purpose 4excludedata
Specify data to be excluded from a fit

Syntax outliers = excludedata(xdata,ydata,’MethodName’,MethodValue)

Arguments xdata A column vector of predictor data.

ydata A column vector of response data.

'MethodName' The data exclusion method.

MethodValue The value associated with MethodName.

outliers A logical vector that defines data to be excluded from a fit.

Description outliers = excludedata(xdata,ydata,’MethodName’,MethodValue)

identifies data to be excluded from a fit using the specified MethodName and
MethodValue. outliers is a logical vector containing 1’s marking data points
to exclude while fitting, and 0’s marking data points to include while fitting.
The data exclusion methods are given below.

Method Description

box A four-element vector that specifies a box of data to include in a

fit. Data outside the box is excluded. You specify the box as
[xmin xmax ymin ymax].

domain A two-element vector that specifies the domain of data to

include in a fit. Data outside this domain is excluded. You
specify the domain as [xmin xmax].

indices A vector specifying the indices of the data points to be excluded.

range A two-element vector that specifies the range of data to include

in a fit. Data outside this range is excluded. You specify the
range as [ymin ymax].

4-98
 excludedata

Remarks You can combine data exclusion methods using logical operators. For example,
to combine methods using the | (OR) operator
outliers = excludedata(xdata,ydata,'indices',[3 5]);
outliers = outliers|excludedata(xdata,ydata,'box',[1 10 0 90]);

In some cases, you might want to use the ~ (NOT) operator to specify a box that
contains all the data to exclude.
outliers = ~excludedata(xdata,ydata,'box',[1 10 0 90]);

Example Generate random data in the interval [0, 15], create a sine wave with noise, and
add two outliers with the value 2.
rand('state',0);
x = 15*rand(150,1);
y = sin(x) + (rand(size(x))-0.5)*0.5;
y(ceil(length(x)*rand(2,1))) = 2;

Identify outliers that are outside the interval [-1.5, 1.5] using the range
method.
outliers = excludedata(x,y,'range',[-1.5 1.5]);

Identify the same outliers using the indices method.

ind = find((y>1.5)|(y<-1.5));
outliers = excludedata(x,y,'indices',ind);

You can pass outliers to the fit function to exclude the specified data points
from a fit.
ftype = fittype('a*sin(b*x)');
fresult = fit(x,y,ftype,'startpoint',[1 1],'exclude',outliers);

See Also fit

4-99
feval

Purpose 4feval
Evaluate a fit result object or a fit type object

Syntax f = feval(fresult,x)
f = feval(ftype,coef1,coef2,...,x)

Arguments fresult A fit result object.

x A column vector of values at which fresult or ftype is

evaluated.

ftype A fit type object.

coef1,coef2,... The model coefficients assigned to ftype.

f A column vector containing the result of evaluating

fresult or ftype at x.

Description f = feval(fresult,x) evaluates the fit result object fresult at the values
specified by x, and returns the result to f. You create a fit result object with the
fit function.

f = feval(ftype,coef1,coef2,...,x) evaluates the fit type object ftype

using the coefficients specified by coef1, coef2, and so on. You create a fit type
object with the fittype function.

Remarks You can also evaluate a fit result or a fit type object using the following syntax.
f = fresult(x);
f = ftype(coef1,coef2,...,x);

Example Create a fit type object and evaluate the object at x using the specified model
coefficients.
x = (0:0.1:10)';
ftype = fittype('a*x^2+b*x');
f = feval(ftype,1,2,x);

4-100
 feval

Create a fit result object and evaluate the object over a finer range in x.
y = x.^2+(rand(size(x))-0.5);
xx = (0:0.05:10)';
fresult = fit(x,y,ftype);
f = feval(fresult,xx);

See Also fit, fittype

4-101
fit

Purpose 4fit
Fit data using a library or custom model, a smoothing spline, or an interpolant

Syntax fresult = fit(xdata,ydata,'ltype')

fresult = fit(xdata,ydata,'ltype','PropertyName',PropertyValue,…)
fresult = fit(xdata,ydata,'ltype',opts)
fresult = fit(xdata,ydata,'ltype',...,'problem',values)
fresult = fit(xdata,ydata,ftype,...)
[fresult,gof] = fit(…)
[fresult,gof,output] = fit(…)

Arguments xdata A column vector of predictor data.

ydata A column vector of response data.

'ltype' The name of a library model, spline, or interpolant.

'PropertyName' The name of a fit options property.

PropertyValue A valid value for PropertyName.

opts A fit options object.

'problem' Specify problem parameters.

values A cell array of problem parameter values.

ftype A fit type object.

fresult The fit result object.

gof Goodness of fit statistics.

output A structure containing information that is associated with

the fitting procedure.

Description fresult = fit(xdata,ydata,'ltype') fits the data specified by xdata and

ydata to the library model, interpolant, or smoothing spline specified by ltype.
The fit result is returned to fresult. You can display the library fit type names
with the cflibhelp function. xdata and ydata cannot contain Infs or NaNs.
Additionally, only the real part of a complex value is used.

4-102
 fit

fresult = fit(xdata,ydata,'ltype','PropertyName',
PropertyValue,...) fits the data using the options specified by PropertyName
and PropertyValue. You can display the fit options available for the specified
library fit type with the fitoptions function.

fresult = fit(xdata,ydata,'ltype',opts) fits the data using options

specified by the fit options object opts. You create a fit options object with the
fitoptions function. This is an alternative syntax to specifying property
name/property value pairs.

fresult = fit(xdata,ydata,'ltype',...,'problem',values) assigns

values to problem parameters. values is a cell array with one element per
parameter. Problem parameters are problem-dependent constants that you
define as part of your model. See fittype for more information on problem
parameters.

fresult = fit(xdata,ydata,ftype,...) fits the data to the fit type object

specified by ftype. You create a fit type object with the fittype function.

[fresult,gof] = fit(...) returns goodness of fit statistics to the structure

gof. The gof structure includes the fields shown below.

Field Description

sse Sum of squares due to error

rsquare Coefficient of determination

dfe Degrees of freedom

adjrsquare Degree-of-freedom adjusted coefficient of determination

rmse Root mean squared error (standard error)

[fresult,gof,output] = fit(...) returns the structure output, which

contains information that is associated with the fitting procedure used.
Supported fitting procedures include linear least squares, robust nonlinear
least squares, and so on. Some information applies to all fitting procedures,
while other information is relevant only for particular fitting procedures. For

4-103
fit

example, the information returned for nonlinear least squares fits is given
below.

Field Description

numobs Number of observations (response values).

numparam Number of unknown parameters to fit.

residuals Vector of residuals.

Jacobian Jacobian matrix.

exitflag Describes the exit condition. If exitflag > 0, the

function converged to a solution. If exitflag = 0, the
maximum number of function evaluations or iterations
was exceeded. If exitflag < 0, the function did not
converge to a solution.

iterations Number of iterations used to complete the fit.

funcCount Number of function evaluations used to complete the fit.

firstorderopt Measure of first-order optimality.

algorithm Fitting algorithm used.

Remarks For rationals and Weibull library models, the coefficient starting values are
randomly selected in the range [0,1]. Therefore, if you perform multiple fits to
a data set using the same equation, you might get different coefficient results
due to different starting values. To avoid this situation, you should pass in a
vector of starting values each time you fit, or define a specific state for the
random number generator, rand or randn, before fitting.
For all other library models, optimal starting points are automatically
calculated. These values depend on the data, and are based on model-specific
heuristics.

4-104
 fit

Example Fit the census data with a second-degree polynomial library model and return
the goodness of fit statistics and the output structure.
load census
[fit1,gof1,out1] = fit(cdate,pop,'poly2');

Normalize the data and fit with a third-degree polynomial.

[fit1,gof1,out1] = fit(cdate,pop,'poly3','Normalize','on');

Fit the data with a single-term exponential library model.

[fit2,gof2,out2] = fit(cdate,pop,'exp1','Normalize','on');

Create a fit options object, and try to find a better fit by overriding the default
starting points for the fit coefficients.
opts = fitoptions('exp1','Norm','on','start',[100 0.1]);
[fit3,gof3,out3] = fit(cdate,pop,'exp1',opts);

Fit the data to a custom model that contains the problem parameter n.
mymodel = fittype('a*exp(b*n*x)+c','problem','n');
opts = fitoptions(mymodel);
set(opts,'normalize','on')
[fit4,gof4,out4] = fit(cdate,pop,mymodel,opts,'problem',{2});

Warning: Start point not provided, choosing random start point.

The warning occurs whenever you fit data with a custom nonlinear model and
do not provide starting points.

See Also cflibhelp, fitoptions, fittype

4-105
fitoptions

Purpose 4fitoptions
Create or modify a fit options object

Syntax opts = fitoptions

opts = fitoptions('ltype')
opts = fitoptions('ltype','PropertyName',PropertyValue,...)
opts = fitoptions('method',value)
opts = fitoptions('method',value,'PropertyName',PropertyValue,...)
opts = fitoptions(opts,'PropertyName',PropertyValue,...)
opts = fitoptions(opts,newopts)

Arguments 'ltype' The name of a library model, spline, or interpolant.

'PropertyName' The name of a fit options property.

PropertyValue A valid value for PropertyName.

’method’ Specify a toolbox fitting method.

value A supported fitting method.

opts,newopts A fit options object.

Description opts = fitoptions creates the empty fit options object opts. The returned
options are supported by all fitting methods, and are given by the following
properties. Note that curly braces denote default property values.

Property Description

Normalize Specifies whether the data is centered and scaled. The

value can be {'off'} or 'on'.

Exclude A vector of one or more data points to exclude from the fit.
You can use the excludedata function to create this vector.

Weights A vector of weights associated with the response data.

Method The fitting method. The value is None for an empty object. A
complete list of supported fitting methods is given below.

4-106
 fitoptions

opts = fitoptions('ltype') creates a default fit options object for the

library or custom fit type specified by ltype. You can display the library model,
interpolant, and smoothing spline names with the cflibhelp function.

opts = fitoptions('ltype','PropertyName',PropertyValue,...) creates

a fit options object for the specified library fit type, and with the specified
property names and property values. Note that you can specify PropertyName
or PropertyValue without regard to case, and you can make use of name
completion by supplying the minimum number of characters that uniquely
identify the string.

opts = fitoptions('method',value) creates a default fit options object for

the fitting method specified by value. A complete list of supported fitting
methods is given below.

opts = fitoptions('method',value,'PropertyName',PropertyValue,...)
creates a default fit options object for the specified fitting method, and with the
specified property names and property values.

opts = fitoptions(opts,'PropertyName',PropertyValue,...) modifies

the existing fit options object with the specified property names and property
values.
opts = fitoptions(opts,newopts) combines the existing fit options object
opts with a new fit options object newopts. If both objects have the same
Method value, the nonempty properties in newopts override the corresponding
properties in opts. If the objects have different Method values, the output object
will have the same Method as opts, and only the Normalize, Exclude, and
Weights properties of newopts will override the corresponding properties in
opts.

Remarks To display the possible fit options property values, use the set function.
set(opts)

To display the current fit options property values, use the get function.
get(opts)

Note that you can configure or display a single property value using the dot
notation. See below for an example.

4-107
fitoptions

Additional Fit Options

If Method is NearestInterpolant, LinearInterpolant, PchipInterpolant, or
CubicSplineInterpolant, there are no additional fit options.

If Method is SmoothingSpline, the SmoothingParam property is available to

configure the smoothing parameter. You can specify any value between 0 and
1. The default value depends on the data set.
If Method is LinearLeastSquares, the additional fit option properties shown
below are available.

Property Description

Robust Specifies whether to use the robust linear least squares

fitting method. The value can be {'off'} or 'on'.

Lower A vector of lower bounds on the coefficients to be fitted. The

default value is an empty vector indicating that the fit is
not constrained by lower bounds. If bounds are specified,
the vector length must equal the number of coefficients. An
unconstrained lower bound is specified by -Inf.

Upper A vector of upper bounds on the coefficients to be fitted. The

default value is an empty vector indicating that the fit is
not constrained by upper bounds. If bounds are specified,
the vector length must equal the number of coefficients. An
unconstrained upper bound is specified by Inf.

4-108
 fitoptions

If Method is NonlinearLeastSquares, the additional fit option properties

shown below are available.

Property Description

Robust Specifies whether to use the robust nonlinear least

squares fitting method. The value can be {'off'} or 'on’.

Lower A vector of lower bounds on the coefficients to be fitted.

The default value is an empty vector indicating that the fit
is not constrained by lower bounds. If bounds are specified,
the vector length must equal the number of coefficients.
An unconstrained lower bound is specified by -Inf.

Upper A vector of upper bounds on the coefficients to be fitted.

The default value is an empty vector indicating that the fit
is not constrained by upper bounds. If bounds are
specified, the vector length must equal the number of
coefficients. An unconstrained upper bound is specified by
Inf.

StartPoint Vector of coefficient starting values. The default value is

an empty vector. If the default value is passed to the fit
function, then starting points for some library models are
determined heuristically. For other models, the values are
selected randomly on the interval (0,1).

Algorithm Algorithm used for the fitting procedure. The value can be
'Levenberg-Marquardt','Gauss-Newton', or
{'Trust-Region'}.

DiffMax Maximum change in coefficients for finite difference

Change gradients. The default value is 0.1.

DiffMin Minimum change in coefficients for finite difference

Change gradients. The default value is 10-8.

Display Level of display. {'notify'} displays output only if the fit

does not converge. 'final' displays only the final output.
'iter' displays output at each iteration. 'off' displays no
output.

4-109
fitoptions

Property Description (Continued)

MaxFunEvals Maximum number of function (model) evaluations

allowed. The default value is 600.

MaxIter Maximum number of fit iterations allowed. The default

value is 400.

TolFun Termination tolerance on the function (model) value. The

default value is 10-6.

TolX Termination tolerance on coefficients. The default value is

10-6.

Example Create an empty fit options object and configure the object so that data is
normalized before fitting.
opts = fitoptions;
opts.Normal = 'on'

opts =

Normalize: 'on'
Exclude: []
Weights: []
Method: 'None'

Creating an empty fit options object is particularly useful when you want to
configure only the Normalize, Exclude, or Weights properties for a data set,
and then fit the data using the same fit options object, but with different fitting
methods. For example, fit the census data using a third-degree polynomial, a
one-term exponential, and a cubic spline.
load census
f1 = fit(cdate,pop,'poly3',opts);
f2 = fit(cdate,pop,'exp1',opts);
f3 = fit(cdate,pop,'cubicsp',opts);

4-110
 fitoptions

You can return values for some fit options with the fit function. For example,
fit the census data using a smoothing spline and return the default smoothing
parameter. Note that this value is based on the data passed to fit.
[f,gof,out] = fit(cdate,pop,'smooth');
smoothparam = out.p
smoothparam =

0.0089

Increase the default smoothing parameter by about 10% and fit again.
opts = fitoptions('Method','Smooth','SmoothingParam',0.0098);
[f,gof,out] = fit(cdate,pop,'smooth',opts);

Create two noisy Gaussian peaks — one with a small width, and one with a
large width.
a1 = 15; b1 = 3; c1 = 0.02;
a2 = 35; b2 = 7.5; c2 = 4;
x = (1:0.01:10)';
rand('state',0)
gdata = a1*exp(-((x-b1)/c1).^2) + a2*exp(-((x-b2)/c2).^2) ...
+ 5*(rand(size(x))-.5);

Fit the data using the two-term Gaussian library model.

ftype = fittype('gauss2');
gfit = fit(x,gdata,ftype);

Maximum number of function evaluations exceeded. Increasing

MaxFunEvals (in fit options) may allow for a better fit, or
the current equation may not be a good model for the data.

Because the Display property is set to its default value Notify, a message is
included as part of the display due to the fit not converging. The message
indicates that you should try increasing the number of function evaluations.

4-111
fitoptions

The fit results are shown below.

gfit
gfit =
General model Gauss2:
gfit(x) = a1*exp(-((x-b1)/c1)^2) + a2*exp(-((x-b2)/c2)^2)
Coefficients (with 95% confidence bounds):
a1 = 43.59 (-411.9, 499.1)
b1 = 7.803 (0.7442, 14.86)
c1 = 4.371 (-3.065, 11.81)
a2 = -10.86 (-373.4, 351.7)
b2 = 11.05 (-190.4, 212.5)
c2 = 6.985 (-124.6, 138.5)

As you can see by examining the fitted coefficients, it is clear that the algorithm
has difficulty fitting the narrow peak, and does a good job fitting the broad
peak. In particular, note that the fitted value of the a2 coefficient is negative.
To help the fitting procedure converge, specify that the lower bounds of the
amplitude and width parameters for both peaks must be greater than zero. To
do this, create a fit options object for the gauss2 model and configure the Lower
property to zero for a1, c1, a2, and c2, but leave b1 and b2 unconstrained.
opts = fitoptions('gauss2');
opts.Lower = [0 -Inf 0 0 -Inf 0];

Fit the data using the new constraints.

gfit = fit(x,gdata,ftype,opts)
gfit =
General model Gauss2:
gfit(x) = a1*exp(-((x-b1)/c1)^2) + a2*exp(-((x-b2)/c2)^2)
Coefficients (with 95% confidence bounds):
a1 = 35 (34.82, 35.17)
b1 = 7.48 (7.455, 7.504)
c1 = 3.993 (3.955, 4.03)
a2 = 4.824 (2.964, 6.684)
b2 = 3 (2.99, 3.01)
c2 = 0.03209 (0.01774, 0.04643)

This is a much better fit, although you can still improve the a2 value.

See Also cflibhelp, fit, get, set

4-112
 fittype

Purpose 4fittype
Create a fit type object

Syntax ftype = fittype('ltype')

ftype = fittype('expr')
ftype = fittype('expr','PropertyName',PropertyValue,...)

Arguments 'ltype' The name of a library model, spline, or interpolant.

'expr' An expression representing a custom model.

'PropertyName' The name of a fit type object property.

PropertyValue A valid value for PropertyName.

ftype A fit type object.

Description ftype = fittype('ltype') creates the fit type object ftype from the library
model, spline, or interpolant specified by ltype. You can display the library fit
type names with the cflibhelp function.

ftype = fittype('expr') creates the fit type object from the expression
specified by expr. The expression expr represents the custom model you will
use to fit your data. To create a general (nonlinear) custom model, specify the
entire equation as one expression. To create a linear custom model, pass in a
cell array of expressions to expr but do not include the coefficients. Each
element of the cell array corresponds to one term of the model. If there is a
constant term, use “1” as the corresponding element in the cell array.
By default, the independent variable is assumed to be x, the dependent
variable is assumed to be y, there are no problem-dependent variables, and all
other variables are assumed to be coefficients of the model. All coefficients
must be scalars.

4-113
fittype

ftype = fittype('expr','PropertyName',PropertyValue,...) creates a fit

type object using the specified property name/property value pairs. The
supported property names are given below.

Property Name Description

coefficients Specify the coefficient names. Use a cell array if there

are multiple names.

dependent Specify the dependent (response) variable name.

independent Specify the independent (predictor) variable name.

options Specify the default fit options for the current expression.

problem Specify the problem-dependent (constant) names. Use a

cell array if there are multiple names.

Example Create a fit type object for a custom general equation and define the
problem-dependent name to be n.
ftype = fittype('a*x+b*exp(n*x)','problem','n');

Define the independent variable to be chan.

ftype = fittype('a*chan+b*exp(n*chan)','ind','chan','prob','n')

ftype =
General model:
ftype(a,b,n,chan) = a*chan+b*exp(n*chan)

Create a fit type object for a custom linear equation and specify names for the
coefficients.
ftype = fittype({'cos(x)','1'},'coeff',{'a1','a2'})

ftype =
Linear model:
ftype(a1,a2,x) = a1*cos(x) + a2

4-114
 fittype

Create a fit type object for the rat33 library model. Note that the display
includes the full equation.
ftype = fittype('rat33')

ftype =
General model Rat33:
ftype(p1,p2,p3,p4,q1,q2,q3,x) = (p1*x^3 + p2*x^2 + p3*x + p4)/
(x^3 + q1*x^2 + q2*x + q3)

Create a fit type object and include the existing fit options object opts, and fit
to the census data.
load census
opts = fitoptions('Method','Nonlinear','Normalize','On');
ftype = fittype('a*exp(b*x)+c','options',opts);
f1 = fit(cdate,pop,ftype);

4-115
get

Purpose 4get
Return properties for a fit options object

Syntax get(opts)
a = get(opts)
a = get(opts,'PropertyName')

Arguments opts A fit options object.

’PropertyName’ The name of a fit options property, or a cell array of

property names.

a A structure or cell array of fit options property values.

Description get(opts) returns all property names and their current values to the
command line for the fit options object opts.

a = get(opts) returns the structure a where each field name is the name of a
property of opts, and each field contains the value of that property.

a = get(opts,'PropertyName') returns the value of the property specified by

PropertyName for opts. If PropertyName is replaced by a cell array of strings
containing property names, get returns a cell array of values to a.

Example Create a fit options object for a second-degree polynomial, and return the
current property values to the command line.
opts = fitoptions('poly2');
get(opts)

ans =
Normalize: 'off'
Exclude: []
Weights: []
Method: 'LinearLeastSquares'
Robust: 'Off'
Lower: []
Upper: []

See Also set

4-116
 integrate

Purpose 4integrate
Integrate a fit result object

Syntax inty = integrate(fresult,x,x0)

Arguments fresult A fit result object.

x The values at which fresult is integrated.

x0 The integration starting point.

inty A vector of integration values.

Description inty = integrate(fresult,x,x0) integrates the fit result object fresult at

the values specified by x starting from x0, and returns the result to inty. The
fresult object is a fit result object generated by the fit function. x is a scalar
or column vector and inty is the same size as x. x0 is a scalar.

Example Create a noisy sine wave on the interval [-2π, 2π].

rand('state',0);
x = (-2*pi:0.1:2*pi)';
y = sin(x) + (rand(size(x))-0.5)*0.2;

Create a custom fit type, and fit the data using reasonable starting values.
ftype = fittype('a*sin(b*x)');
fit1 = fit(x,y,ftype,'startpoint',[1 1]);

Calculate the integral for each value of x starting at -2*pi

inty = integrate(fit1,x,x(1));

See Also differentiate, cfit, fit, quad

4-117
plot

Purpose 4plot
Plot data, fit, prediction bounds, outliers, and residuals

Syntax plot(fresult)
plot(fresult,xdata,ydata)
plot(fresult,xdata,ydata,'s')
plot(fresult,'s1',xdata,ydata,'s2')
plot(fresult,xdata,ydata,outliers)
plot(fresult,xdata,ydata,outliers,'s')
plot(...,'ptype1','ptype2',...)
plot(...,'ptype1','ptype2',...,conflev)
h = plot(…)

Arguments fresult A fit result object.

xdata A column vector of predictor data.

ydata A column vector of response data.

s,s1,s2 The plot symbols, plot colors, and line type.

outliers A vector of outliers.

'ptype' The plot type. You can specify multiple plot types as a cell
array of strings.

conflev The confidence level.

h A vector of plot handles.

Description plot(fresult) plots the fit result object fresult. fresult is a fit result object
generated by the fit function.

plot(fresult,xdata,ydata) plots the fit result object, the predictor data

specified by xdata, and the response data specified by ydata.

plot(fresult,xdata,ydata,'s') plots the predictor and response data using

the color, symbol, and line type specified by the string s. Refer to the built-in
plot function for color, symbol, and line type options.

4-118
 plot

plot(fresult,'s1',xdata,ydata,'s2') plots the fit result object using the

color, symbol, and line type specified by the string s1, and plots the predictor
and response data using the color, symbol, and line type specified by the string
s2.

plot(fresult,xdata,ydata,outliers) plots the outliers specified by

outliers in a different color. outliers must be the same size as xdata and
ydata. You identify data points as outliers with the excludedata function.

plot(fresult,xdata,ydata,outliers,'s') plots the outliers using the color,

symbol, and line type specified by the string s.

plot(...,'ptype1','ptype2',...) plots the plot types specified by ptype1,

ptype2, and so on. ptype can be a single plot type or multiple plot types, which
you can specify as a cell array of strings. For one plot type or none (the default),
plot behaves like the built-in plot command and draws into the current figure
and axes. This way, you can use commands like subplot and hold to arrange
plots in a figure window and to superimpose multiple fits into the same graph.
For multiple plot types, plot uses subplot to create one set of axes per plot
type. The supported plot types are given below.

Plot Type Description

fit Plot the data and the fit (default).

predfunc Same as fit but with prediction bounds for the function.

predobs Same as fit but with prediction bounds for a new

observation.

residuals Plot the residuals. The fit corresponds to the zero line.

stresiduals Plot the standardized residuals. The fit corresponds to the

zero line. Standardized residuals are the ordinary residuals
divided by their standard deviation. Standardizing puts all
residuals on a common scale (units of standard deviations)
and makes it easier to quantify how far a point is from the
fitted curve.

4-119
plot

plot(...,'ptype1','ptype2',...,conflev) plots prediction bounds with

the confidence level specified by conflev. conflev must be between 0 and 1.
The default value is 0.95 for 95% confidence levels.

h = plot(…) returns a vector of handles to h.

Remarks To plot error bars, use the errorbar function. For example, if you have a vector
of weights w (reciprocal variances) associated with the response data ydata, you
can plot symmetric error bars with the following command.
errorbar(xdata,ydata,1./sqrt(w))

Example Create a noisy sine wave on the interval [-2π, 2π] and add two outliers with the
value 2.
rand('state',2);
x = (-2*pi:0.1:2*pi)';
y = sin(x) + (rand(size(x))-0.5)*0.2;
y(ceil(length(x)*rand(2,1))) = 2;

Identify outliers that are outside the interval [-1.5, 1.5] using the range
method.
outliers = excludedata(x,y,'range',[-1.5 1.5]);

Create a custom fit type, define fit options that exclude the outliers from the fit
and define reasonable starting values, and fit the data.
ftype = fittype('a*sin(b*x)');
opts = fitoptions('Method','NonLinear','excl',outliers,...
'Start',[1 1]);
fit1 = fit(x,y,ftype,opts);

Plot the data, the fit to the data, and mark the outliers.
subplot(2,1,1)
plot(fit1,'k-',x,y,'b.',outliers,'ro');

4-120
 plot

Plot the residuals.

subplot(2,1,2)
plot(fit1,'k-',x,y,'b.','residuals');

2
data
1.5 excluded data
fitted curve
1

0.5
y

−0.5

−1

−1.5
−8 −6 −4 −2 0 2 4 6 8
x

2
data
zero line
1.5
y residual

0.5

−0.5
−8 −6 −4 −2 0 2 4 6 8
x

4-121
plot

Plot 99% confidence and prediction bounds for the function and for a new
observation.
plot(fit1,'k-',x,y,'b.','predf','predo',0.99);

2
data
1.5 fitted curve
confidence bounds
1

0.5
y

−0.5

−1

−1.5
−8 −6 −4 −2 0 2 4 6 8
x

2
data
1.5 fitted curve
confidence bounds
1

0.5
y

−0.5

−1

−1.5
−8 −6 −4 −2 0 2 4 6 8
x

See Also errorbar, plot (built-in), subplot

4-122
 predint

Purpose 4predint
Compute prediction bounds for new observations or for the function

Syntax ci = predint(fresult,x)
ci = predint(fresult,x,level)
ci = predint(fresult,x,level,'intopt','simopt')
[ci,ypred] = predint(...)

Arguments fresult A fit result object.

x The values at which predictions are calculated.

level Confidence level. The value must be between 0 and 1. The

default value is 0.95.

'intopt' Can be observation (the default) to compute bounds for new

response values, or functional to compute bounds for the fit
evaluated at x.

'simopt' Can be off (the default) to compute nonsimultaneous

bounds, or on to compute simultaneous bounds.

ci An array of upper and lower prediction bounds.

ypred The predicted (fitted) value of fresult evaluated at x.

Description ci = predint(fresult,x) returns prediction bounds for new response

(observation) values at the predictor values specified by x. The confidence level
of the predictions is 95%. ci contains the upper and lower prediction bounds.
fresult is the fit result object returned by the fit function. You can compute
prediction bounds only for parametric fits. To compute confidence bounds for
the fitted parameters, use the confint function.

ci = predint(fresult,x,level) returns prediction bounds with a confidence

level specified by level.

ci = predint(fitresult,x,level,'intopt','simopt') specifies the type of

bounds to compute. If intopt is functional, the bounds measure the
uncertainty in estimating the function (the fitted curve). If intopt is
observation, the bounds are wider to represent the additional uncertainty in
predicting a new response value (the fitted curve plus random noise).

4-123
predint

If simopt is off, nonsimultaneous bounds are calculated. If simopt is on,

simultaneous bounds are calculated. Nonsimultaneous bounds take into
account only individual x values. Simultaneous bounds take into account all x
values.

[ci,ypred] = predint(...) returns the predicted (fitted) value of fresult

evaluated at x.

Example Generate some data and add noise.

x = (0:0.2:10)';
coef = [2 -0.2];
rand('state',0)
y = coef(1)*exp(coef(2)*x) + (rand(size(x))-0.5)*0.5;

Fit the data using a single-term exponential and define the range over which
prediction bounds are calculated.
fresult = fit(x,y,'exp1');

Return the prediction bounds for the function as well as the predicted values of
the fit using nonsimultaneous and simultaneous bounds with a 95% confidence
level. For nonsimultaneous bounds, given a single predetermined predictor
value, you have 95% confidence that the true function lies between the
confidence bounds. For simultaneous bounds, you have 95% confidence that the
function at all predictor values lies between the bounds.
[c1,ypred1] = predint(fresult,x,0.95,'fun','off');
[c2,ypred2] = predint(fresult,x,0.95,'fun','on');

Return the prediction bounds for new observations as well as the predicted
values of the fit using nonsimultaneous and simultaneous bounds with a 95%
confidence level. For nonsimultaneous bounds, given a single predictor value,
you have 95% confidence that a new observation lies between the confidence
bounds. For simultaneous bounds, regardless of the predictor value, you have
95% confidence that a new observation lies between the bounds.
[c3,ypred3] = predint(fresult,x,0.95,'obs','off');
[c4,ypred4] = predint(fresult,x,0.95,'obs','on');

4-124
 predint

Plot the data, fit, and confidence bounds.

subplot(2,2,1), plot(fresult,x,y), hold on, plot(x,c1,'k-.')
legend('data','fitted curve','prediction bounds')
title('Nonsimultaneous bounds for function')
subplot(2,2,3), plot(fresult,x,y), hold on, plot(x,c2,'k-.')
legend('data','fitted curve','prediction bounds')
title('Simultaneous bounds for function')
subplot(2,2,2), plot(fresult,x,y), hold on; plot(x,c3,'k-.')
legend('data','fitted curve','prediction bounds')
title('Nonsimultaneous bounds for observation')
subplot(2,2,4), plot(fresult,x,y), hold on, plot(x,c4,'k-.')
legend('data','fitted curve','prediction bounds')
title('Simultaneous bounds for observation')

Nonsimultaneous bounds for function Nonsimultaneous bounds for observation

2.5 2.5
data data
fitted curve 2 fitted curve
2 prediction bounds prediction bounds
1.5
1.5
1
y

1
0.5
0.5 0

0 −0.5
0 2 4 6 8 10 0 2 4 6 8 10
x x

Simultaneous bounds for function Simultaneous bounds for observation

2.5 2.5
data data
fitted curve 2 fitted curve
2 prediction bounds prediction bounds
1.5
1.5
1
y

1
0.5
0.5 0

0 −0.5
0 2 4 6 8 10 0 2 4 6 8 10
x x

See Also confint, fit

4-125
set

Purpose 4set
Configure or display property values for a fit options object

Syntax set(opts)
a = set(opts)
set(opts,'PropertyName',PropertyValue,...)
set(opts,PN,PV)
set(opts,S)

Arguments opts A fit options object.

’PropertyName’ A property name for opts.

PropertyValue A property value supported by PropertyName.

PN A cell array of property names.

PV A cell array of property values.

S A structure with property names and property values.

a A structure array whose field names are the property

names for opts, or cell array of possible values.

Description set(opts) displays all configurable property values for the fit options object
opts. If a property has a finite list of possible string values, these values are
also displayed.

a = set(opts) returns all configurable properties and their possible values

for opts to the structure a. The field names of a are the property names of opts,
and the field values are cell arrays of possible property values. If the property
does not have a finite set of possible values, the cell array is empty.

set(opts,'PropertyName',PropertyValue,...) configures multiple

property values with a single command.

set(opts,PN,PV) configures the properties specified in the cell array of strings

PN to the corresponding values in the cell array PV.

set(opts,S) configures the named properties to the specified values for opts.
The structure S has field names given by the fit options object properties, and
the field values are the values of the corresponding properties.

4-126
 set

Example Create a custom nonlinear model, and create a default fit options object for the
model.
mymodel = fittype('a*x^2+b*exp(n*c*x)','prob','n');
opts = fitoptions(mymodel);

Configure the Robust and Normalize properties using property name/property

value pairs.
set(opts,'Robust','LAR','Normalize','On')

Configure the Display, Lower, and Algorithm properties using cell arrays of
property names and property values.
set(opts,{'Disp','Low','Alg'},{'Final',[0 0 0],'Levenberg'})

See Also fitoptions, get

4-127
smooth

Purpose 4smooth
Smooth the response data

Syntax yy = smooth(ydata)
yy = smooth(ydata,span)
yy = smooth(ydata,'method')
yy = smooth(ydata,span,'method')
yy = smooth(ydata,'sgolay',degree)
yy = smooth(ydata,span,'sgolay',degree)
yy = smooth(xdata,ydata,...)

Arguments ydata A column vector of response data.

span The number of data points to include for each smooth

calculation.

'method' The smoothing method.

’sgolay’ Use Savitzky-Golay smoothing.

degree The polynomial degree for the Savitzky-Golay method.

xdata A column vector of predictor data.

yy A vector of smoothed response data.

Description yy = smooth(ydata) smooths the response data specified by ydata using the
moving average method. The default number of data points in the average (the
span) is five. yy is the smoothed response data. Note that you need not specify
the predictor data if it is sorted and uniform.

yy = smooth(ydata,span) uses the number of data points specified by span in

the moving average calculation. span must be odd.

yy = smooth(ydata,'method') smooths the response data using the method

specified by method and the default span. The supported smoothing methods

4-128
 smooth

are given below. For the Savitzky-Golay method, the default polynomial degree
is 2.

Method Description

moving Moving average filter.

lowess Locally weighted scatter plot smooth using least squares

linear polynomial fitting.

loess Locally weighted scatter plot smooth using least squares

quadratic polynomial fitting.

sgolay Savitzky-Golay filter. Note that the algorithm used by the

toolbox can accept nonuniform predictor data.

rlowess Lowess smoothing that is resistant to outliers.

rloess Loess smoothing that is resistant to outliers.

yy = smooth(ydata,span,'method') smooths data using the specified span

and method. For the loess and lowess methods, you can specify span as a
percentage of the total number of data points. In this case, span must be less
than or equal to 1. For the moving average and Savitzky-Golay methods, span
must be odd. If an even span is specified, it is reduced by 1.

yy = smooth(ydata,'sgolay',degree) uses the Savitzky-Golay method with

polynomial degree specified by degree.

yy = smooth(ydata,span,'sgolay',degree) uses the number of data points

specified by span in the Savitzky-Golay calculation. span must be odd and
degree must be less than span.

yy = smooth(xdata,ydata,...) smooths the data specified by ydata and the

associated predictor data, xdata. You should specify the predictor data when it
is not uniformly spaced or it is not sorted. If xdata is not uniform and you do
not specify method, lowess is used. If the smoothing method requires xdata to
be sorted, the sorting occurs automatically.

4-129
smooth

Remarks For the moving average and Savitzky-Golay methods, span must be odd. If an
even span is specified, it is reduced by 1. If span is greater than the length of
ydata, it is reduced to the length of ydata.

Use robust smoothing when you want to assign lower weight to outliers. The
robust smoothing algorithm uses the 6MAD method, which assigns zero weight
to data outside six mean absolute deviations.
Another way to generate a vector of smoothed response values is to fit your
data using a smoothing spline. Refer to the fit function for more information.

Example Suppose you want to smooth traffic count data with a moving average filter to
see the average traffic flow over a 5-hour window (span is 5).
load count.dat
y = count(:,1);
yy = smooth(y);

Plot the original data and the smoothed data.

t = 1:length(y);
plot(t,y,'r-.',t,yy,'b-')
legend('Original Data','Smoothed Data Using ''moving''',2)

120
Original Data
Smoothed Data Using ’moving’

100

80

60

40

20

0
0 5 10 15 20 25

4-130
 smooth

The first four elements of yy are given by

yy(1) = y(1)
yy(2) = (y(1)+y(2)+y(3))/3
yy(3) = (y(1)+y(2)+y(3)+y(4)+y(5))/5
yy(4) = (y(2)+y(3)+y(4)+y(5)+y(6))/5

Because of the way that the end points are treated, the result shown above
differs from the result returned by the filter function described in “Difference
Equations and Filtering” in the MATLAB documentation.
In this example, generate random data between 0 and 15, create a sine wave
with noise, and add two outliers with the value 3.
rand('state',2);
x = 15*rand(150,1);
y = sin(x) + (rand(size(x))-0.5)*0.5;
y(ceil(length(x)*rand(2,1))) = 3;

Smooth the data using the loess and rloess methods with the span specified
as 10% of the data.
yy1 = smooth(x,y,0.1,'loess');
yy2 = smooth(x,y,0.1,'rloess');

Plot original data and the smoothed data.

[xx,ind] = sort(x);
subplot(2,1,1)
plot(xx,y(ind),'r.',xx,yy1(ind),'k-')
set(gca,'YLim',[-1.5 3.5])
legend('Original Data','Smoothed Data Using ''loess''',2)
subplot(2,1,2)
plot(xx,y(ind),'r.',xx,yy2(ind),'k-')
set(gca,'YLim',[-1.5 3.5])
legend('Original Data','Smoothed Data Using ''rloess''',2)

4-131
smooth

Note how the outliers have less effect with the robust method.

3
Original Data
2 Smoothed Data Using ’loess’

−1

0 5 10 15

3
Original Data
2 Smoothed Data Using ’rloess’

−1

0 5 10 15

See Also fit, sort

4-132
 Index

Numerics
6MAD center and scale 1-8
robust fitting 3-12 cfit function 4-81
robust smoothing 2-17 cflibhelp function 4-82
cftool function 4-84
coefficient
A confidence bounds 3-33
adjusted residuals 3-12 constraints
adjusted R-square 3-31 Fit Options GUI 3-23
algorithms 3-15 Fourier series example 3-53
Analysis GUI Gaussian example 3-59
census data example 1-15 starting values
description 4-89 Fit Options GUI 3-23
axes limit control Gaussian example 3-59
census data example 1-10 structure
nonparametric fit example 3-74 piecewise polynomials 3-72
coefficient of multiple determination 3-30
complex data
B importing 2-3
backslash operator 3-9 confidence bounds
batch mode 1-19 census data example 1-13
best fit 1-8 definition 3-32
bisquare weights Legendre polynomial example 3-50
robust fitting 3-11 confint function 4-90
robust smoothing 2-17 constraints
bounds Fit Options GUI 3-23
confidence Fourier series example 3-53
census data example 1-13 Gaussian example 3-59
definition 3-32 covariance matrix of coefficient estimates 3-34
prediction Create Custom Equation GUI
definition 3-32 definition 3-21
goodness of fit example 3-37 Legendre polynomial example 3-46
cubic spline interpolation 3-68
curve fitting session
C saving custom equations 3-20
carbon12alpha data set 3-46
saving fit results 1-18
census data example 1-3

I-1
 Index

Curve Fitting Tool Smooth pane 2-10

Fourier series example 3-54 data sets
Gaussian example 3-58 deleting 2-3
Legendre polynomial example 3-50 enso 3-52
nonparametric fit example 3-74 flvote2k 3-61
opening with cftool 4-84 gauss3 3-57
prediction bounds 3-39 hahn1 3-41
rational example 3-43 importing 2-2
residuals 3-37 renaming 2-3
robust fit example 3-64 Data Sets pane
starting 1-2 census data example 1-4
custom equations description 2-2
definition 3-20 data tips
general accessing 2-6
Fourier series example 3-52 robust fit example 3-64
Gaussian example 3-57 datastats function 4-92
robust fit example 3-61 default
linear coefficient parameters
Legendre polynomial example 3-46 fit options 3-26
robust fitting example 3-61 confidence level for bounds 3-33
saving 3-20 smoothing parameter 3-70
degrees of freedom 3-31
deleting
D data sets 2-3
data 3-61 exclusion rules 2-26
excluding 2-25 design matrix 3-8
fitting procedure determining the best fit 1-8
census data example 1-5 differentiate function 4-94
general steps 3-2 disp function 4-96
importing 1-3
quality 3-5
sectioning 2-25 E
smoothing 2-9 enso data set 3-52
statistics 4-92 equations
See also predictor data, response data custom 3-20
Data GUI library 3-16
Data Sets pane 2-2 error distribution 3-5

I-2
 Index

error sum of squares 3-30 moving average 2-12

evaluating the goodness of fit Savitzky-Golay 2-19
available measures 3-27 finite differencing parameters 3-24
example 3-37 fit convergence criteria 3-24
examples Fit Editor
evaluating the goodness of fit 3-37 census data example 1-5
excluding and sectioning data 2-32 Legendre polynomial example 3-48
Fourier series fit 3-52 nonparametric fit example 3-72
Gaussian fit 3-57 robust fit example 3-64
importing data 2-4 fit function 4-102
Legendre polynomial fit 3-46 Fit Options GUI
nonparametric fit 3-72 description 3-23
rational fit 3-41 Fourier series example 3-53
robust fit 3-61 Gaussian example 3-59
sectioning periodic data 2-35 fit tips 2-6
smoothing data 2-21 fitoptions function 4-106
Exclude GUI fitting
description 2-26 algorithms 3-15
example 2-34 batch mode 1-19
robust fit example 3-63 fit options 3-23
excludedata function 4-98 least squares method
excluding data definition 3-6
example 2-32 linear 3-6
marking outliers 2-27 nonlinear 3-14
sectioning 2-30 nonparametric 3-68
exclusion rule numerical results 1-11
definition 2-25 parametric 3-4
robust fitting example 3-63 procedure
exponentials census data example 1-5
fit type definition 3-16 general steps 3-2
extrapolation visual results 1-8
census data example 1-15 Fitting GUI
census data example 1-6
Fit Editor 1-5
F fitting process 3-2
feval function 4-100 nonparametric fit example 3-72
filtering data numerical fit results 1-11

I-3
 Index

robust fit example 3-64 description 3-22

Table of Fits 1-5 Fourier series example 3-52
fittype function 4-113 Gaussian example 3-57
flvote2k data set 3-61 generating an M-file 1-19
Fourier series get function 4-116
example 3-52 goodness of fit
fit type definition 3-16 available measures 3-27
functions census data example 1-8
cfit 4-81 statistics 3-29
cflibhelp 4-82 graphically viewing data 2-6
cftool 4-84 GUI
confint 4-90 Analysis
datastats 4-92 census data example 1-15
differentiate 4-94 description 4-89
disp 4-96 Create Custom Equation
excludedata 4-98 definition 3-21
feval 4-100 Legendre polynomial example 3-46
fit 4-102 Data 2-2
fitoptions 4-106 Exclude 2-26
fittype 4-113 Fit Options
get 4-116 description 3-23
integrate 4-117 Fourier series example 3-53
plot 4-118 Gaussian example 3-59
predint 4-123 Fitting
set 4-126 census data example 1-6
smooth 4-128 Legendre polynomial example 3-48
nonparametric fit example 3-72
rational example 3-42
G robust fit example 3-64
gauss3 data set 3-57 Plotting
Gaussian census data example 1-17
error distribution 3-5 description 4-88
example 3-57 smoothing data example 2-23
fit type definition 3-17 Table Options
Gauss-Newton algorithm 3-15 census data example 1-11
general equations 3-22 goodness of fit evaluation 3-38
General Equations pane

I-4
 Index

H linear equations
hahn1 data set 3-41 custom 3-21
hat matrix 3-9 fit options 3-23
fitting 3-6
Linear Equations pane
I description 3-21
importing data 1-3 Legendre polynomial example 3-49
description 2-2 robust fit example 3-62
example 2-4 linear interpolation 3-68
influential data 2-28 linear least squares 3-6
Infs loading the curve-fitting session 1-18
importing 2-3 local regression smoothing 2-14
removing 2-41 loess 2-14
integrate function 4-117 lowess 2-14
interpolants 3-68
iteratively reweighted least squares 3-12
M
MAD
J robust fitting 3-12
Jacobian 3-14 robust smoothing 2-17
marking outliers 2-27
median absolute deviation
L robust fitting 3-12
LAR 3-11 robust smoothing 2-17
least absolute residuals 3-11 M-file generation 1-19
least squares fitting models
definition 3-6 custom 3-20
linear 3-6 library 3-16
nonlinear 3-14 moving average filtering 2-12
robust 3-11 multiple correlation coefficient 3-30
weighted linear 3-9
Legendre polynomials
example 3-46 N
generating 3-47 NaNs
Levenberg-Marquardt algorithm 3-15 importing 2-3
leverages 3-12 removing 2-41
library models 3-16 nearest neighbor interpolation 3-68

I-5
 Index

nonlinear equations description 4-88

custom 3-22 smoothing data example 2-23
fit options 3-23 polynomials
fitting 3-14 census data example 1-5
nonlinear least squares 3-14 fit type definition 3-17
nonparametric fitting Legendre 3-47
example 3-72 piecewise 3-70
methods 3-68 rational models 3-19
normal distribution 3-5 power series
normal equations 3-7 fit type definition 3-18
normalization 1-8 prediction bounds
numerically viewing data 2-8 definition 3-32
goodness of fit example 3-37
predictor data
O census data example 1-3
objects importing 2-3
cfit 4-81 sectioning 2-27
fit result 4-102 viewing numerically 2-8
fittype 4-113 predint function 4-123
outliers preprocessing data 2-41
definition 2-25 excluding and sectioning 2-25
marking 2-27 smoothing 2-9
removing 2-41 transforming the response data 2-40
robust fit 3-11 Preview window 2-5
overdetermined system of equations 3-7 problem parameter
overfitting passing to fit 4-105
census data example 1-13 projection matrix 3-9
goodness of fit evaluation 3-39

Q
P QR decomposition 3-9
parametric fitting 3-4 quality of data
pchip 3-70 definition 3-5
piecewise polynomials 3-70 weighted linear least squares 3-9
plot function 4-118
Plotting GUI
census data example 1-17

I-6
 Index

R robust least squares 3-11

rationals robust weights
example 3-41 fitting 3-12
fit type definition 3-19 smoothing 2-17
regression Rodrigues’ formula 3-47
sum of squares 3-30 root mean squared error 3-31
weights R-square 3-30
least squares 3-9 adjusted 3-31
smoothing 2-14 negative values 3-31
removing Infs, NaNs, and outliers 2-41 rubber band selection 2-33
renaming
data sets 2-3
exclusion rules 2-26 S
residual degrees of freedom 3-31 saving
residuals analysis results
adjusted 3-12 workspace variables 1-16
comparing multiple fits 1-8 fit results
definition 3-27 curve fitting session 1-18
displaying M-file 1-19
census data example 1-7 workspace variables 1-13
goodness of fit evaluation 3-27 Savitzky-Golay filtering 2-19
excluding data with 2-33 scatter plot 2-6
response data scatter plot smooth 2-14
census data example 1-3 sectioning data
error distribution 3-5 definition 2-25
importing 2-3 example 2-32
sectioning 2-27 periodic data example 2-35
transforming 2-40 rules 2-30
viewing numerically 2-8 session 1-18
right-click menu 2-6 set function 4-126
RMSE 3-31 shape-preserving interpolation 3-68
robust sine functions 3-19
fitting smooth function 4-128
alternative to excluding data 2-29 Smooth pane
example 3-61 description 2-10
regression schemes 3-11 example 2-22
smoothing 2-17

I-7
 Index

smoothing data tricube weights 2-14

definition 2-9 trust-region algorithm 3-15
example 2-21
local regression 2-14
moving average filtering 2-12 V
robust procedure 2-17 variances 3-11
Savitzky-Golay filtering 2-19 viewing data
smoothing spline 3-70 graphically 2-6
span 2-9 numerically 2-8
spline
cubic interpolant 3-68
smoothing 3-70 W
spline 3-70 Weibull distribution
SSE, SSR, SST 3-30 fit type definition 3-20
standard error 3-31 weighted linear least squares 3-9
standardized residuals 4-119 weights 3-11
starting values regression
Fit Options GUI 3-23 least squares 3-9
Gaussian example 3-59 smoothing 2-14
structure of coefficients 3-72 robust
sum of sine functions least squares 3-11
fit type definition 3-19 smoothing 2-17
sum of squares 3-6 viewing numerically 2-8
error 3-30
regression 3-30
total 3-30

T
Table of Fits 1-5
Table Options GUI
census data example 1-11
goodness of fit evaluation 3-38
toolbar 2-6
Tools menu 2-6
total sum of squares 3-30
transforming the response data 2-40

I-8