E G A R E G R E S S I O N A N A L Y S I S


By

Paul A. Wangerin
1821 W. Daisy Lane
Glendale, Wisconsin 53209
(414) 351-6254




EGA Regression Analysis is an easy to use program designed for
simple 2-variable correlation analysis. Its main purpose is to take
advantage of the properties of EGA graphics in conducting such an
analysis. The user of this program is free to inspect and test it at
no charge. If you should find this program of some utility a donation
of $5.00 sent to the author would be greatly appreciated. You will
then be placed on a mailing list and informed of any future revisions.

PLEASE feel free to send any comments and suggestions-- a response
will be given.

I hope you enjoy this program!


USING THE PROGRAM: (Entering the data and setting the parameters)
-----------------

NUMBER OF DATA POINTS--Whether data is entered from the keyboard
or from a file you must input the number of X,Y points to be used.

SIZE OF SCATTER POINTS--A number between 1 (smallest) to 3
(largest) is entered here. Generally the larger the data set the
smaller you would want your plotted points to be and vice versa.
Experiment with this option to your own satisfaction.

FILE CREATION AND STRUCTURE--Data files are created using a pure
ASCII editor such as DOS EDLIN and have the following structure:

"URBAN GROWTH" (Name of X-variable in quotes)
"##.##" (Print using FORMAT for values on X-axis of graph)
"POP. GROWTH RATE" (Name of Y-variable in quotes)
"###.#" (Print using FORMAT for values on Y-axis of graph)
2.33,5
4.22,10
3.16,7.5 X,Y pairs (Separated by a Comma)
-1.2,-3
6.4,12.22
etc.

Note: Sample files have been included with the program called
SAMPLE1.DAT (N=14), SAMPLE2.DAT (N=39) and SAMPLE3.DAT (N=11).








-Page 2-


You may also input your values from the keyboard--simply select
this option when prompted. Note that the maximum length of the format
string is 5 characters. Also, you do not need to use quotes for
variable names and formats when entering them from the keyboard.
DEFAULT values may be selected by hitting . Values for the X,Y
points are then entered one at a time when prompted.

DATA TRANSFORMATION:

Data may be transformed to account for skewness in the marginal
distributions and in effect to reduce non-linear equations to their
linear form. Three basic transformations for the x-variable are
utilized: Log base 10, reciprocal (1/x), and x-squared. The effect
of the first two transformations is to skew the x-distribution in
the negative direction while the x-squared transformation will
skew the distribution in the positive direction.

NOTES ON THE OUTPUT:

1. The red-line on the graph represents the LEAST-SQUARES regression
line of Y predicted from X.

2. The vertical red-line represents the size of the Standard Error of
Y from Y predicted:
2
( ( Y' - Y ) / N )

3. Y is always the dependent variable, X the independent or predictor
variable.

4. Slope and intercept represent the values of b and a in the
equation of the form: Y' = a + bX

5. Significance is one minus the probability of ZERO correlation
in the population.


PRINTING THE SCREEN:

In order to print the screen you need to have installed an EGA
SCREEN PRINT utility. DOS GRAPHICS command will NOT print screens of
this resolution. The author of this program has found Charles
Petzold's EGAGRAF.COM (from PC-MAGAZINE "DOS Power House Tools") very
effective in this regard.




This program was written and compiled using MICROSOFT QUICK BASIC 4.0.

* * *