0% found this document useful (0 votes)
61 views13 pages

User Manual: Section of This Document, Beginning On Page 5, For Details of How To Use These Procedures

The document describes RegressIt, an Excel add-in for regression analysis that features a 40-button ribbon interface controlling analysis generation and navigation, with versions for both PC and Mac platforms. RegressIt adds a ribbon to Excel containing buttons organized into groups for analysis, model space navigation, worksheet navigation, cell display, and utilities. The buttons launch procedures for descriptive statistics, regression modeling, and navigating between output worksheets.

Uploaded by

TheGimhan123
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
61 views13 pages

User Manual: Section of This Document, Beginning On Page 5, For Details of How To Use These Procedures

The document describes RegressIt, an Excel add-in for regression analysis that features a 40-button ribbon interface controlling analysis generation and navigation, with versions for both PC and Mac platforms. RegressIt adds a ribbon to Excel containing buttons organized into groups for analysis, model space navigation, worksheet navigation, cell display, and utilities. The buttons launch procedures for descriptive statistics, regression modeling, and navigating between output worksheets.

Uploaded by

TheGimhan123
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 13

USER MANUAL1

RegressItTM is an Excel add-in for multivariate descriptive data analysis and regression analysis, first released to the
public in 2014 at regressit.com. The 2018 version features a novel 40-button ribbon interface which controls every
aspect of the generation and navigation and presentation of your data analysis, as well as providing access to
internal documentation and teaching materials. It is not necessary to use the native Excel ribbon or scrolling and
paging tools nor do any typing during a RegressIt session apart from optional customization of analysis names. The
ribbon interface is designed so that the entire analysis can be driven efficiently not only with a mouse but also by
tapping one’s finger on a touchscreen, and the output files are easily viewable and navigable on smart phones.
There are separate linear regression versions of the program for PC’s and Macs, and they have identical features.
A separate logistic regression version runs on PC’s This document illustrates the use of the linear regression version.

RegressIt can be launched either by opening its file as needed or by installing it on the Excel add-ins menu. See the
installation instructions on the web site for details. When it is running, it adds a new ribbon to Excel’s top menu.
The buttons on the ribbon are organized into 6 main groups, whose options are described below.

ANALYSIS

Descriptive Statistics: launches the descriptive stats procedure


Linear Regression: launches the regression procedure

From the dialog boxes of both procedures, you can launch the variable transformation procedure. See the next
section of this document, beginning on page 5, for details of how to use these procedures.

MODEL SPACE NAVIGATION

VCR buttons: move one worksheet to the right or left, or jump to data sheet (leftmost sheet) or model summary
sheet (rightmost sheet)
Last stats: toggle back and forth to the last-visited stats worksheet from wherever you are.
Last model: toggle back and forth to the last-visited model worksheet from wherever you are.

1
February 21, 2018. Additional documentation for the R interface is contained in a separate document.
1
Note: these buttons only work if you use RegressIt’s navigation tools to move around in the workbook, not if
you jump to other worksheets by clicking on their tabs at the bottom of the screen,
History: activates a dialog box with a list of all worksheets in the file and identifying information about the analysis
on each one. You can select one and jump directly to it from here. The list remains open while you do this
so you can keep jumping around. This also provides a handy view of the audit trail: great for browsing
through files produced by yourself or others to see what was done at what date and time on what computer.
Relatives: navigates through a sequence of regression models by parent-child relationships
See page 9 for examples of using the History and Relatives tools.
Compare: Hitting this button while positioned on the title row of any table or chart will reposition it at the top of
the viewing area and do the same for all other model sheets, so that the same pieces of output will appear
in the same relative positions on the screen when moving back and forth among model sheets to compare
their results. Also, this tool together with the Top and Zoom tools can be used to re-set the entire
workbook to a pristine viewing state, with cell A1 in the top left corner of the window, possibly with a
different choice of zoom level to suit a new computer (e.g., for a presentation).
Filter: this executes the Excel filter tool. It can be used to filter and sort the coefficient table (say, on P-values) or
to sort a table of squared correlations with the dependent variable (to pre-select independent variables) or
to sort the residual table on error size or other stats there.
Remove: clicking this button while selecting some range of cells in the coefficient table for an existing regression
model will cause those rows to be grayed-out, which de-selects them from the next regression model
launched from that sheet. This is intended to be used in conjunction with the filter tool: sort the coefficient
table on P-values and deselect a bunch of insignificant variables all at once. It can also be used on a
descriptive statistics output sheet in which the “squared correlations with first variable” option was chosen.
If you launch a regression model from that sheet, the variables listed there will be picked up by default
unless some have been deselected with this tool. This is convenient for throwing out variables with
insignificant correlations with the dependent variable when going from descriptive analysis to regression
analysis, particularly when the set of candidate variables is very large.
Export R code: exports R code for the regression model on whose sheet you are positioned. There is a menu of
options for table and chart output. If chart output is chosen, the chart titles in R include the model name
from RegressIt. After hitting this button and clicking through the options, go to the command prompt in
RStudio and hit <Paste><Enter> to run the code. You can also copy R code for regression models from R to
RegressIt: just copy and paste the model specification (“…<-lm(…)” ) text string onto a cell in the Excel file
and launch the regression procedure from that cell. See the RegressIt R code interface document for details.

WORKSHEET NAVIGATION

Up/down: jump up and down by whole table or whole chart, so that the uppermost one is always correctly aligned
with the top left corner of the screen. This is important for efficient viewing because it doesn’t require fine
adjustment of the viewing area with the scroll bar. It also minimizes re-drawing speed, which is important
for hi-res or editable charts. This feature is also very useful in live presentations.
Zoom in/out: zooms all worksheets at once in a series of predefined levels.
Top: jump to top left of current worksheet
Show all: toggle the display of all charts and tables at once by grouping or ungrouping of their rows. When they
are toggled off while title rows remain displayed, you just see a list of all tables and charts on the sheet, and
they can be selectively displayed from there.
2
Show Tables: toggle the display of tables only
Show Charts: toggle the display of charts only
Title rows: toggle title rows on and off. Above each table or chart is a title row that includes the same model details
(name, dep. variable, sample size) that are in chart titles. For audit trail purposes it is useful to include this
extra line in the copied area when tables are individually copied to other documents.
Conf +/- : these buttons interactively increase or decrease the confidence level for confidence limits shown in tables
and editable charts on the regression worksheet. (Fun for demonstrations.)

CELL DISPLAY

Colors: toggle fill colors on and off. This tool affects correlations and autocorrelations on the descriptive stats
sheet, t-stats and standardized coefficients and correlations of coefficients on the model sheet, and
coefficients in the model summary worksheet. There is a predefined scale of fill colors ranging from dark
blue (very significantly positive) to dark red (very significantly negative).
Fonts: toggle font coding on and off. This tool affects the same cells as the coloring tool, but changes the color
(shade of gray) and boldness of font rather than fill color as an indicator of significance (better for printing).
Notes: toggles red flags for cell comments on and off. When the flags are on, the comment will pop up next to the
cell when you hover the mouse over it. There is always a cell comment behind the model name on the
model sheet and behind the run time cell on the model summary sheet, containing more facts about the
model (e.g., run time and name of user and file at run time, for audit trail purposes). The coefficient cells
on the model summary worksheet also have comments with more details of each coefficient (t-stats,
stamdardized values, etc.). On the model sheet, if the “teaching notes in cell comments” box was checked
at run time, there are teaching notes behind the titles in the regression statistics table and coefficient table
(R-squared and all that). Customization of the teaching notes to fit an instructor’s own materials is
something that could be done.
View: Hitting this key while positioned in a cell that contains a comment (whether or not its flag is visible) will cause
the comment to be displayed at the top center of the Excel window. This is the best way to view comments
when the Excel window is not very wide, because otherwise the comments may extend off the right of the
screen Also, this is the only way to view comments on a touchscreen or tablet computer when not using a
mouse, because hovering over a cell is not possible in that case.

UTILITIES

Select data: if you hit this button while positioned anywhere on the data worksheet, it will feel out and select the
range containing the data.
Create names: executes the create-name-from-selection command to assign variable names (= range names). Hit
this immediately after hitting the “Select data” button when setting up a new file.
Export data: exports the contents of the data worksheet to a CSV file and writes R code to the clipboard for loading
it into RStudio. There is an option to customize the file name and location when doing this. The default file
name is the data worksheet name and default file location is the current file location. After hitting this

3
button, go to the command prompt in RStudio and hit <Paste><Enter> to run the code, which loads the
data into a frame with the given name.
Gridlines: toggles gridlines and column headings on and off. This can be used for manually re-sizing or hiding rows
and columns. (Normally this is not needed: the output is designed for easy reading by default.)
Toolbar H: launches the horizontal floating toolbar interface (below), which has the same button layout as the
ribbon but can be positioned anywhere on the computer screen (even outside the Excel window),
something else that can be useful for presentations. It is better than the ribbon for use on a touchscreen
or tablet computer. You can put it at the bottom of the screen so that your hand won’t block your view of
the spreadsheet. The buttons are also a bit bigger than those on the ribbon, better for clumsy fingers.

Toolbar V: launches the vertical floating toolbar interface, which can be positioned to the left or right of the Excel
window as shown below. This is generally the best interface to use on a touchscreen or tablet computer
because it leaves more vertical space for output while keeping your hand off to one side. It has even bigger
buttons. This toolbar can also be activated by hitting Ctrl-G.

SUPPORT: These buttons at the right of the ribbon display message boxes with general
comments on how to use the program, as well as program version information.
You should check the website (regressit.com) periodically for updates, particularly
if you have encountered technical issues.

4
PROCEDURE MENUS AND SAMPLE OUTPUT

DESCRIPTIVE ANALYSIS PROCEDURE: This procedure always produces a descriptive stats table and correlation
matrix. There are also options for autocorrelations, series plots, scatterplots, and histogram plots. See pages 10 to
13 of this handout for screen shots of RegressIt output worksheets. There is an option to choose the variable to be
listed first, i.e. topmost in the stats table and correlation matrix and first in the ordering of charts that are produced.
This is normally used when one of the variables is pre-identified as the dependent variable to be used in subsequent
regression analysis. If the “squared correlations with first variable” option is chosen, the correlation matrix is
replaced by a table of correlations and squared correlations versus the first variable, as a prelude to regression.
Regression variables can be pre-selected from it by using the “Remove” tool on the ribbon to deselect those having
only tiny correlations with the prospective dependent variable.

There are 3 graph format options. The choice of which one to use generally depends on sample size and
presentation quality that is needed. It is possible to have a very large number of separate charts on the output
worksheet (hundreds or even thousands—no kidding), although drawing speed and file size are limiting factors,
particularly with large sample sizes. Editable graphs require much less storage space than pictures for small sample
sizes, vice versa for big ones. However, they are continuously re-drawn when the worksheet is scrolled up and
down, which can be problematic for large data sets, particularly on a Mac. The Up/Down tools on the RegressIt
toolbar help to mitigate this issue, because they move up or down by whole chart, keeping the topmost one in
alignment, without the need to fine-tune their vertical position using the scroll bar. Storage space for hi-res pictures
also depends on sample size. They are fixed format, but they are not bitmaps: they contain the data, although you
5
can’t touch it. Bottom line: for large sample sizes (many thousands) it is generally best to go with low-res picture
format. It looks fine unless the graph is enlarged significantly (e.g. to fill a whole Powerpoint slide). Hi-res picture
and editable charts remain razor-sharp under enlargement and are more efficient for smaller sample sizes. Another
consideration in chart type selection is that on low-res and high-res charts, text and graphic elements scale up
proportionally under enlargement, while on editable charts they don’t.

LAGS FOR AUTOCORRELATIONS: This option applies to both descriptive statistics (autocorrelations of variables)
and regression (autocorrelations of residuals). Autocorrelations can be consecutive or not. For example, for
monthly time series data, you might choose only lag 1 and lag 12.

LINEAR REGRESSION PROCEDURE:

If you click on a plus sign, it will open a message box with a description of the model option to the right.

6
Standard text output (summary stats table, coefficient table, ANOVA table) is always produced, and there are many
options for additional tables and charts. Forecasts can be generated for rows where the dependent variable is
missing and all independent variables are present. Teaching notes that explain the key output statistics can be
optionally embedded in the output worksheet as cell comments.

Model names for use in table and chart titles can be customized. If the name of the previous model was “12-pack
model 3.2”, the default for the next one will be “12-pack model 3.3,” but it can be changed arbitrarily before
running the model. The model name entered in this field will also become the output worksheet name.

The model information row in table and chart titles can also be fully customized if needed, independently of the
worksheet name. The dialog box below appears next if the “Customize titles” box is checked. This feature is useful
when preparing a final set of charts and tables for presentation or inclusion in a report. Students can be also asked
to include identifying information about themselves or their team.

For example, when this custom title is entered…

…the charts come out looking like this:

However, it is generally best to use the default chart titles for routine work, because they link back to the names of
the worksheets in the file from which the charts were copied. It’s usually enough to just be thoughtful in assigning
model names, and RegressIt’s default naming convention helps with this.

Point sizes on graphs are automatically scaled to suit the size of the data set.

7
VARIABLE TRANSFORMATION TOOL: This can be launched from either the descriptive stats or regression dialog
box. It includes time transformations if the “Time Series Data” box is checked when it is launched. Transformed
variables are automatically assigned descriptive names, e.g., the natural log of X is automatically named X.LN. X
lagged by 1 period is automatically named X.LAG1, and so on. The transformation name is tacked on at the end
rather than the beginning so that the original and transformed variables appear adjacent to each other in
alphabetically sorted variable lists. Also, the new column of data for the transformed variable is inserted next to
the column of the parent variable on the data worksheet.

8
HISTORY TOOL: Provides a list of all worksheets in the file and allows you to jump directly to any of them. This box
stays open on the screen until “Cancel” is hit, so you can visit a sequence of worksheets without having to re-launch
it. On a PC it can be kept on top of the worksheet while you work. On a Mac you should position it outside the
Excel window. The last sheet visited is remembered for use by the “Last Stats” and “Last Models” buttons, so that
you can easily toggle back and forth between any two sheets.

Audit trail information for regression models, such as run time and computer name, is stored here and in several
other places in the workbook, both in text strings in cells and in cell comments. This is useful for managing your
own individual work or teamwork by keeping track of the temporal sequence as well as which of your many
computers you may have used to store related files. It also makes it hard for a student to hand in a file created by
someone else for a homework assignment or a laboratory exercise. More detailed information about all regression
models in the analysis sequence (error stats, coefficients, etc.) is stored on the Model Summaries worksheet in
publication format, i.e. side-by-side columns that allow easy comparisons of model parameters. Information about
all models previously created remains there even if their model worksheets are later deleted, whereas the history
list is dynamically created to show only worksheets currently in the file.

RELATIVES TOOL: Allows you to jump to parents or children of the regression model whose sheet you are currently
looking at, in order to easily retrace your steps or back up to an earlier model in order to go down another search
path. (How does it work? Each regression model worksheet contains information about its parents and children,
hidden in white font. Look in cells Q2, R2, etc.)

9
Sample output worksheet from the
DESCRIPTIVE STATISTICS procedure, showing
descriptive stats, autocorrelations, series plots,
histogram plots, correlations, and scatterplots.

Here, the series plots include optional trend


lines and their slopes as well as an optional
custom time axis scale.

Scatterplots include optional regression lines


and center-of-mass points.

Color-coding has been turned on for the


autocorrelations and correlations. This can be
toggled on and off at any time if RegressIt is
active.

The editable option was used, so that all 24


charts are native Excel charts that are subject
to editing, although their descriptive titles and
default formatting generally make this
unnecessary.

10
Sample output worksheet from the LINEAR REGRESSION
procedure, including forecasts that were automatically
generated for 4 missing values of the dependent variable
at the end of the series.

Color coding has been toggled on for the t-stats,


standardized coefficients, and residual autocorrelations.
It would also be applied to the correlation matrix of
coefficients if that had been included.

The residual table, if it had been included, would appear


at the bottom and would include actual and predicted
values, residuals, standardized residuals and their
absolute values, leverage, and Cook’s D. The filter tool on
the ribbon can be used to easily sort the residual table on
any of these statistics.

11
TEACHING NOTES can optionally be
included as cell comments behind the
key statistics in the output.

They can be viewed by hovering the


mouse over the cell (when its flag is
turned on) or by clicking on the cell and
then clicking the View button on the
ribbon. The latter method will cause
them to be displayed at the top center
of the Excel window, rather than off to
the side, better for small screens.

12
MODEL SUMMARIES WORKSHEET

Example of a model summaries worksheet, which shows side-by-side side model stats in journal-article
format. Optional color coding of coefficients has been toggled on here too.

This sheet (which is located to the right of all analysis worksheets) is updated with every new regression
model, and it provides an audit trail for all models that have been fitted in this file’s lifetime, even those
whose output worksheets may have been subsequently deleted. It is very easy to spin off large numbers
of models with memory-intensive chart output, so the user will often want delete inferior ones in the
course of a long analysis. However, it is good to maintain the entire audit trail, particularly to verify the
efforts of students.

The cell comment behind the model name includes much more audit trail information, and cell comments
behind the coefficient estimates provide more detail there.

The user has the option at run time to store the coefficients in the table in coefficient plus P-value text
format (which was done here), or else in numeric form, either standardized or unstandardized. Regardless
of which option was chosen for display, the cell comments provide all the other information about each
coefficient estimate.

Another feature is that any regression model whose summary output is stored here can be easily re-run.
The model specifications are stored in the cell right above the model name and hidden in white font. If
the regression button is hit while positioned on this cell, the model information will be read from it, and
that model’s worksheet will be moved to if it is present. If it is not present, the variable specifications will
become the default selections for a new model, so the original one can be regenerated with a couple of
clicks. The entire history of models in the workbook can be re-created from this single sheet in this way.

If models had also been fitted to some other dependent variable, their results would be stored in a
different block of rows farther down. Thus, only models fitted to the same dependent variable are
displayed side-by-side.

13

You might also like