Chapter 3: Curve Fitting – Contents

– Introduction
• Functions of Curve Fitting
• Types of Curve Fitting
– Linear Regression
• Quantification of Error of Linear Regression
– Linearization of Non-linear Models
– Polynomial Regression
– Interpolation
• Linear Interpolation
• Quadratic Interpolation
• Cubic Interpolation 1
 Functions of Curve Fitting
1. You may require estimates at points b/n the data of
discrete values.
– techniques to fit curves to such data to obtain intermediate
estimates.
2. To make a simplified version of a complicated
function.
– compute values of the function at a number of discrete values
along the range of interest.
– then, a simpler function may be derived to fit these values.
• Both of these applications are known as Curve Fitting.

2
 Types of Curve Fitting
• There are two general approaches for curve fitting based
on the amount of error associated with the data.
1. Where the data exhibits a significant error, derive a
single curve that represents the general trend of the
data.
– Because any individual data point may be incorrect, we make
no effort to intersect every point.
– Rather, the curve is designed to follow the pattern of the
points taken as a group.
– One approach of this nature is called least-squares regression
(Fig. 3.1a).
2. Where the data is known to be very precise, fit a curve
that passes directly through each of the points. 3
 Types of Curve Fitting
– The estimation of values between well-known discrete points
is called Interpolation (Fig. 3.1b and c).

(a) Linear Regression: did not attempt to connect the

points; followed the general upward trend of the data
with a straight-line.
(b) Linear Interpolation: used straight-line segments to
connect the points (very common practice in eng’g).
(c) Curvilinear Interpolation: used curves to try to
4
capture the data.
 Least-squares Regression
• Linear Regression: how to fit the “best” straight-line
through a set of uncertain data points.
• Some criterion must be devised to establish a basis for the fit
• Choose a straight-line that minimizes the discrepancy b/n the
data points and the line
• Minimize the sum of the squares of the residuals b/n the
measured y and the y calculated with the linear model

…. (a)
• This criterion yields a unique line for a given set of data.
• To determine values for a0 and a1, Eq. (a) is differentiated wrt
each coefficient:
… (b)
5
 Least-squares Regression
• All summations are from i = 1 to n.
• Setting the derivatives equal to zero will result in a
minimum Sr.

…. (c)
• Since we can express the equations as a set of
two simultaneous linear equations with two unknowns
(a0 & a1):
…. (d)
The equations can be solved simultaneously as,
and … (e, f)
6
 Example 3.1: Linear Regression
• Fit a straight-line to the x and y values in the table
below.
x y
1 0.5
2 2.5
3 2.0
4 4.0
5 3.5
6 6.0
7 5.5

7
 Example 3.1: Linear Regression
• Soln.

8
 Example 3.1: Linear Regression

• Any line other than the one computed in

Example 3.1 results in a larger sum of
the squares of the residuals. 9
 Quantification of Error of Linear Regression
• Use analogy with Statistics:
– Central Tendency – mean, median, mode
– Dispersion of Data – Standard Deviation, Coefficient of Variation.

10
 Quantification of Error of Linear Regression
• The analogy can be extended to the determination of “Standard
Deviation” for the regression line as:

11
 Improvement Due to Linear Regression
• Improvement

12
 Improvement Due to Linear Regression
• Improvement

13
Example 3.2: Estimation of error for the linear least-squares fit
• Compute the total standard deviation, the standard error of the
estimate, and the correlation coefficient for the data in Example
3.1.

14
Example 3.2: Estimation of error for the linear least-squares fit
• Example 3.2

15
 Example 3.3: Linear Regression
• Fit a straight line to the x and y values in the table
below. Compute the Total Standard Deviation, Standard
Error of Estimate and the Coefficient of Determination.

x y
10 25
20 70
30 380
40 550
50 610
60 1220
70 830
80 1450 16
 Example 3.3: Linear Regression
• Soln.

17
 Linearization of Non-linear Relationships
• Linear regression provides a powerful technique for
fitting a best line to data.
• However, it is predicated on the fact that the relationship
b/n the dependent and independent variables is linear.
• This is not always the case, and the first step in any
regression analysis should be to plot and visually inspect
the data to ascertain whether a linear model applies.

• Fig. 3.7 (a) Data ill-suited for linear least-squares regression, (b) Parabola is preferable.
18
 Linearization of Non-linear Relationships
• Exponential Model

19
 Linearization of Non-linear Relationships
• Power Eqn. and Saturation-growth-rate Eqn.

20
 Linearization of Non-linear Relationships
• Exponential Eqn. and Power Eqn. Transformations

21
 Linearization of Non-linear Relationships
• Saturation-growth-rate Eqn. Transformations

22
 Linearization of Non-linear Relationships
• Exponential Eqn. and Power Eqn. Transformations

23
 Linearization of Non-linear Relationships
• Saturation-growth-rate Eqn. Transformations

24
Example 3.4: Linearization of Non-linear Eqns.
• Fit a power equation to the data given below by using
linear transformation

x y
1 0.5
2 1.7
3 3.4
4 5.7
5 8.4

25
Example 3.4: Linearization of Non-linear Eqns.
• Soln

26
Example 3.4: Linearization of Non-linear Eqns.
• Soln

27
Example 3.5: Linearization of Non-linear Eqns.
• A

28
Example 3.5: Linearization of Non-linear Eqns.
• Soln.

29
Example 3.5: Linearization of Non-linear Eqns.
• Soln.

30
 Polynomial Regression
• PR

31
 Polynomial Regression
• PR

32
 Polynomial Regression
• PR

33
 Example 3.6: Polynomial Regression
• Fit a 2nd – order polynomial (m = 2) to the data given
below.
x y
0 2.1
1 7.7
2 13.6
3 27.2
4 40.9
5 61.1

34
 Example 3.6: Polynomial Regression
• Soln.

35
 Example 3.6: Polynomial Regression
• Soln.

36
 Interpolation
• Interpolation can be used to:
– estimate intermediate values between precise data points;
– approximate a complicated function by a polynomial function which is
simple to differentiate and integrate
• For n + 1 data points, there is one and only one polynomial of
order n that passes through all the points.
– there is only one straight line that connects two points
– only one parabola connects a set of three points
• Polynomial Interpolation consists of determining a unique nth-
order polynomial that fits n + 1 data points.
• The polynomial then provides a formula to compute intermediate
values.

37
 Interpolation
• there are a variety of mathematical formats in which nth-order
polynomial can be expressed;
• two alternatives that are well-suited for computer implementation:
Newton and Lagrange Interpolating Polynomials.
• Newton Interpolating Polynomials: Linear Interpolation

• The notation f1(x) designates that this is

a first-order interpolating polynomial.
• : this term is the slope of the
line and a finite-divided-difference
approximation of 1st derivative
38
 Example 3.7: Linear Interpolation
• Estimate the natural logarithm of 2 using linear interpolation.
First, perform the computation by interpolating b/n ln 1 = 0 and
ln 6 = 1.791759. Then, repeat the procedure, but use a smaller
interval from ln 1 = 0 to ln 4 = 1.386294.
• Note that the true value of ln 2 ≈ 0.6931472.
Solution:
f(x) = ln x, where x = 2, x0 = 1, x1 = 6

εt = 48.3%.
• Using smaller interval from x0 = 1 to x1 = 4 yields
εt = 33.3%.

• Thus, using the smaller interval reduces the % relative error, εt .

39
 Example 3.7: Linear Interpolation
• Two linear interpolations to estimate ln 2

40
 Quadratic Interpolation
• For more accurate results, introduce some curvature into the line
connecting the points
• For three data points, use a 2nd-order polynomial
• One convenient form for this purpose is
f2(x) = b0 + b1(x − x0) + b2(x − x0)(x − x1)
f2(x) = b0 + b1x − b1x0 + b2x2 + b2x0x1 − b2xx0 − b2xx1
or, collecting terms,
f2(x) = a0 + a1x + a2x2
• a0 = b0 − b1x0 + b2x0x1
• a1 = b1 − b2x0 − b2x1
• a2 = b2
• Recall that the general form of nth-order polynomial is:
f(x) = a0 + a1x + a2x2 +· · ·+anxn 41
 Quadratic Interpolation
• A simple procedure can be used to determine the values of the
coefficients, b0 , b1, b2.
• b0 can be computed with x = x0 → b0 = f(x0)
• b1 can be computed with x = x1 →

• b2 can be computed with x = x2 →

• b1 still represents the slope of the line connecting points x0 and x1.
• The last term, b2(x − x0)(x − x1), introduces the 2nd -order
curvature into the formula.

42
 Example 3.8: Quadratic Interpolation
• Fit a 2nd-order polynomial to the three points used in Example 3.7
• x0 = 1 f(x0) = 0
• x1 = 4 f(x1) = 1.386294
• x2 = 6 f(x2) = 1.791759
• Use the polynomial to evaluate ln 2.
Solution
• b0 = f(x0) = 0

• f2(x) = 0 + 0.4620981(x − 1) − 0.0518731(x − 1)(x − 4)

• f2(2) = 0.5658444, εt = 18.4%. (TV = ln 2 ≈ 0.6931472.)
43
 General Form of Newton’s Interpolating Polynomials
• General form to fit an nth-order polynomial to n + 1 data points
fn(x) = b0 + b1(x − x0)+· · ·+bn(x − x0)(x − x1) · · · (x − xn−1)
• Use data points [x0, f (x0)], [x1, f (x1)], . . . , [xn, f (xn)] and the
following equations to evaluate the coefficients:
• b0 = f(x0)
• b1 = f [x1, x0]
• b2 = f [x2, x1, x0]
• bn = f [xn, xn−1, . . . , x1, x0]
• the bracketed function evaluations are finite divided differences.

• the nth finite divided difference is

44
 General Form of Newton’s Interpolating Polynomials
• The interpolating polynomial is
fn(x) = f (x0) + (x − x0) f [x1, x0] + (x − x0)(x − x1) f [x2, x1, x0]
+ · · ·+(x − x0)(x − x1) · · · (x − xn−1) f [xn, xn−1, . . . , x0]
• which is called Newton’s divided-difference interpolating
polynomial.
• It is not necessary that the data points be equally spaced or that
the abscissa values necessarily be in ascending order.
• Also, notice how the Eqs. are recursive—that is, higher-order
differences are computed by taking differences of lower-order
differences (see Fig. below).

45
 Example 3.9: NIP
• In Example 3.8, data points at x0 = 1, x1 = 4, and x2 = 6 were used
to estimate ln 2 with a parabola. Now, adding a fourth point
[x3 = 5; f (x3) = 1.609438], estimate ln 2 with a third-order
Newton’s Interpolating Polynomial.
Solution
• The 3rd-order polynomial with n = 3, is
f3(x) = b0 + b1(x − x0) + b2(x − x0)(x − x1) + b3(x − x0)(x − x1)(x − x2)
• The first divided differences for the problem are

46
 Example 3.9 Solution: NIP
• The second divided differences are

• The third divided difference is (with n = 3)

• The results for f [x1, x0], f [x2, x1, x0], and f [x3, x2, x1, x0] represent
the coefficients b1, b2, and b3, respectively.
• Along with b0 = f (x0) = 0.0, the cubic Eqn. is
f3(x) = 0 + 0.4620981(x − 1) − 0.05187311(x − 1)(x − 4)
+ 0.007865529(x − 1)(x − 4)(x − 6)
• which can be used to evaluate f3(2) = 0.6287686, εt = 9.3%.
47
 Example 3.9 Solution: NIP
• The complete cubic polynomial is shown in Figure below.

The use of cubic interpolation to estimate ln 2

48
 Lagrange Interpolating Polynomials
• The Lagrange Interpolating Polynomial is simply a reformulation
of the Newton polynomial that avoids the computation of divided
differences. It can be represented concisely as
• where

• where П designates the “product of.” For example, the linear

version (n = 1) is

• and the second-order version is

49
 Lagrange Interpolating Polynomials
• Each term Li (x) will be 1 at x = xi and 0 at all other sample points.
• Thus, each product Li(x) f(xi) takes on the value of f(xi) at the
sample point xi.
• Consequently, the summation of all the products designated by
fn(x) is the unique nth order polynomial that passes exactly
through all n + 1 data points.

50
 Example 3.10: Lagrange Interpolating Polynomials
• Use a Lagrange interpolating polynomial of the first and second
order to evaluate ln 2, on the basis of the data given in Ex. 3.8:
• x0 = 1 f(x0) = 0
• x1 = 4 f(x1) = 1.386294
• x2 = 6 f(x2) = 1.791760
Solution
• The 1st-order polynomial can be used to estimate at x = 2,

• The second-order polynomial is developed as

• As expected, both these results agree with those previously 51

obtained using Newton’s interpolating polynomial.