PD ROM 1/PD ROM Volume I - Macintosh Software from BMUG (1988).iso/Education/Mathematics/Multi Reg/MultiReg Help

Text File | 1987-03-31 | 9.3 KB | 459 lines | [TEXT/ttxt]

1

MultiReg is a program designed to perform Multiple
Linear Regression via the classic least squares analysis.
By the use of this technique, you can fit complicated
mathematical equations to a set of points. MultiReg is
complete with a set of tests which allow you to ensure
that the proper assumptions are satisfied.
Written by Bob James, GEnie BOB.JAMES,
Compuserve 72325,1615.

/////

2

This is the main menu.  You should choose one of the
subprograms to perform.  If a choice is dimmed, that
means you cannot choose it at this time.  For example,
when MultiReg first starts, you can only create a new
problem, get an old problem from the disk, or quit.
Some other limitations are: you can't remove a variable
if you only have one; you can't remove an observation
if you already are at the minimum level; and you can
perform the K-S normality test with fewer than 6
observations.

/////

3

MultiReg is waiting for you to choose one of the
subprograms.  Please click on one of the active (dark)
buttons.

/////

4

MultiReg is waiting for you to click the button.

/////

5

You are now in the initialization portion of MultiReg.
After this is completed, you will be presented with the
main menu.

/////

6

MultiReg is waiting for your input.

/////

7

Don't do anything - MultiReg is doing the work.

/////

8

Click either the YES or the NO button

/////

10

This subprogram is used when you want to enter a new
problem.  You will be asked to enter the number of
decision variables, their names, and the numerical
value for each observation.  These data are not saved
to disk until you choose the Disk File Management
subprogram.

/////

11

Tell MultiReg how many independent variables you
have  (how many "factors" are affecting your
Y-values).  You must have at least 1, and no more than
configured in the initialization process.

/////

12

You must give a name to each of the variables.  Just
type in some name (1-10 characters).

/////

13

For this observations what is the value of the
dependent (Y) variable?

/////

14

For this observation, please enter the values of the
independent variables.  MultiReg is prompting you with
the name of the variable; you should enter some
numerical value.

/////

20

This subprogram will solve the current problem.  After
the solution is complete, you will automatically be
transferred to the display of the estimated
paremeters.

/////

30

This subprogram is used for a display of the raw data
as you entered or retrieved from disk.  The format is as
currently specified in the Display menu.

/////

40

This subprogram displays the estimated model
parameters as computed by the MultiReg program.  The
"Sig" column is starred ("*") if the parameter is
significantly different from zero at the alpha/2
significance level, as currently specified in the Alpha
menu.

/////

50

This subprogram allows you to retrieve an old problem
from disk, save your current problem to a new or old
disk file, or delete any existing file from disk.  The
"Update" button allows you to read a file from disk,
make changes, and re-write it quickly to the disk.  It is
just like saving the current problem under the same
name it was when you retrieved it.  If a button is
dimmed, you cannot perform that action now.  Choose
"Cancel" to return to the main menu without doing
anything.

/////

51

Choose one of the dark buttons.  If you don't want to
perform any disk operation, choose "Cancel."

/////

52

Enter any file name.  Your current problem will be
saved as that name.

/////

53

Choose one of the files for reading.

/////

60

This subprogram shows the two-sided confidence limits
according to the alpha level currently specified in the
Alpha menu.  If the confidence interval does not cross
zero, the "Sig" column will be starred.

/////

70

This subprogram lets you change any existing numerical
information.  In order to do this, you need to know the
index of each observation (they are numbered in the
order you originally entered them into the computer).
If you need to find out the index of the observation
you want to change, use the "Display Data"
subprogram.

/////

71

Enter an observation number to delete.  If you aren't
sure, just press the <RETURN> key.  Observation
numbers are available via the "Display Data"
subprogram.

/////

80

This is the classic Analysis of Variance for significance
of regression.  If your model is significant at or below
the alpha level currently specified in the Alpha menu,
that will be noted on the output (a message printed
right after the "Probability of Chance" entry).

/////

90

With this subprogram, you can use your fitted model to
predict a Y-value based on values of the independent
variables.  You must enter these values; the predicted
Y-value is displayed, along with a prediction an
confidence interval on the mean, both based on the
alpha level currently specified under the Alpha menu.

/////

91

Enter the values for each independent variable as you
are prompted.

/////

100

This subprogram will allow you to list the residual
values (the difference between the predicted Y-value
and the observed Y-value from your data).  Also, you
can generate a plot of any two sets of data, chosen
from the Y-values, any of the independent variables,
the residuals, or the index numbers of each
observation.

/////

101

Choose one of the listed variables for the data set to
be used for the axis.

/////

110

Use this subprogram to remove any one or more of the
observations.

/////

111

Enter the number of the observation which you want
removed.  If you don't want to remove any, just press
the <RETURN> key.

/////

120

This subprogram will allow you to transform any
variable with one of the transformations displayed
when you chose the subprogram.  If the transformation
won't work, MultiReg will let you know.  If it will, you
will need to choose a name for the new variable (or use
the one MultiReg suggests).  Some of the
transformations need further information, and will
request it.

/////

121

Choose the variable which you want transformed.  This
variable will be retained, and a new one created.

/////

122

Choose one of the transformations.  If you have
changed your mind, just click the "Cancel" button.

/////

Type in a new variable name, or just click the OK button
to use the one MultiReg has selected for you.

/////

130

This subprogram lets you delete any variable and all of
its associated observations.

/////

131

Select the variable you want to delete.  If you have
changed your mind, click the "Cancel" button.

/////

140

This subprogram will compute the correlation
coefficient for all possible combinations of Y-values
and dependent variables.  You should watch for high
multicollinearity between dependent variables, which
could cause numerical problems in computations.

/////

150

The Kolgomorov-Smirnov test checks the residuals to
see if they can be assumed to be normally distributed
about zero.  This is an assumption of multiple linear
regression.  Your problem should pass this test (that is,
the assumption of normality should be accepted).  Don't
forget to perform the test of independence also.

/////

160

This subprogram lets you add more observations to the
current problem.

/////

161

Enter the Y-value for this new observation.  If you are
finished, just press the <RETURN> key.

/////

170

This subprogram tests the assumption that the
residuals are independent, by performing the
Durbin-Watson check for first-order corrrelation.  Your
data should pass this test.  This test is only
approximate if you have fewer than 15 observations or
more than 5 dependent variables.

/////

180

You are now quitting MultiReg.

/////

181

You can return to the main menu by clicking on the
"No" button; otherwise, you will quit MultiReg.

/////

190

If you get this message, you have entered into the
depths of the MultiReg wizards area!

/////

200

This routine is used to allow you to create a sample
data file which can be used by MultiReg. You can
specify the number of parameters, the variable
counts, coefficients, and values, and the standard
error. The actual X-values and error terms will
be generated randomly.

/////

201

Tell MultiReg how many independent variables
(X-variables) you want. They will be named
X1, X2, X3, etc. You must have at least 1 and
no more than 100. You are not bound by the
limit of the number of variables in your current
execution of MultiReg.

/////

202

How many sample points do you want generated?
You must have at least 2 more than the number of
independent variables.

/////

203

Enter the constant term of your equation which
relates Y with the X-variables. This is sometimes
known as the intercept.

/////

204

The values of the independent variables (X's) will
be generated as integers by MultiReg. You get to
specify the range of values (low and high).

/////

205

If you want, you can supply the coefficients of the
independent variables. If you choose not to supply
them, the coefficient of X1 will be set to 1, the
coefficient of X2 to 2, and so forth.

/////

206

MultiReg is waiting for you to enter the coefficient
of the stated independent variable.

/////

207

Enter the standard deviation of the error term.
MultiReg will generate an error for each
observation, which will be a normally distributed
number with a mean of zero and a standard
deviation as you specify.

/////

208

Don't do anything. MultiReg is creating the
sample data file and saving it on the disk
file you specified earlier.

/////