USING MINITAB FOR REGRESSION

Say that the data for Y is contained in C2 and the data for X is contained in C1

To make a scatterplot use the plot command.

MTB > PLOT C2*C1 (always PLOT Y*X)

A graphics window will open which contains the scatterplot. At the top of the window are three boxes.

Left hand box -

If you click on the left hand box the window will be minimized; that is, it will disappear but it is still accessible. To activate the window again, click on Windows at the top of your screen. A menu will appear which shows all the windows that are currently open in your project. There is a check by the window which is currently active. To activate any window in the project click on that window.

Middle box -

If you click on the middle box the graph window will be maximized; that is, it will fill the screen.

Right hand box -

If you click on the right hand box (the one with the X in it), the window will be deleted from your project.

If you wish to print a graphics window, activate the window that you wish to print and click on Print Graph in the File menu.

To get regression output for simple linear regression

MTB > REGRESS C2 on 1 in C1 (C2 = y, C1 = x)

To get the residuals, use the subcommand RESIDUALS and specify a column for them.

MTB > REGRESS C2 on 1 in C1; (C2=y, C1=x)

SUBC > RESIDUALS C3.( Now the residuals are stored in C3)

When you press enter, you will get the normal regression output. However, the residuals

are stored in C3. Look at column 3 in the worksheet.

If you want to plot the residuals vs x,

MTB > PLOT C3 *C1 (residuals in C3, x in C1)

If you want to find out what the predicted value of y is at some value of x, say x=4

use the PREDICT subcommand

MTB > REGRESS C2 on 1 in C1;

SUBC > PREDICT 4.

At the end of the regular regression output you will find information on the predicted value.

You can stack subcommands so if you want to make several predictions, say at x=2, x=5.5, x=12, and also store the residuals in C3, type

MTB > REGRESS C2 on 1 in C1;

SUBC > PREDICT 2;

SUBC > PREDICT 5.5;

SUBC > PREDICT 12;

SUBC > RESIDUALS C3.

(Note that at the end of the last subcommand, you need a period. If you forget the period and get an error message, just type in a period and press enter.)

If you wish to find the correlation between two variables, use the command CORR. The commands CORR C1 C2 and CORR C2 C1 will give the same value since the correlation between x and y is the same as the correlation between y and x.

Note that you can use any column numbers you want to, just make sure to use column numbers that you haven't used before or you will overwrite the original columns.

IF YOU ARE USING MINITAB 7 on gs

The commands are the same except for the PLOT command

MTB > PLOT C2 C1 (no star is used for the plot command)

You get the residuals in the same way

MTB > REGRESS C2 on 1 in C1; (C2=y, C1=x)

SUBC > RESIDUALS C3.( Now the residuals are stored in C3)

When you press enter, you will get the normal regression output but you will not see the residuals. They are stored in C3; to access them, at the MTB prompt, type PRINT C3 and press return.

FOR BOTH VERSION 12 AND VERSION 7 OF MINITAB

NOTE: I have used capital letters to indicate the commands: however, MINITAB is not case sensitive, so you may use lower case letters for both the commands and the columns i.e. describe c1 works just as well as DESCRIBE C1.

Jonathan Payne
January 15, 1999