Random Sampler version 1.52

CONTENTS
	Program Overview
	Installing Random Sampler
	Changes from prior version

Program Overview

Random Sampler is a program for conducting Monte Carlo 
analyses on sample continuous variable data that is 
subdivided according to a categorical variable.  Some
examples are vehicle miles per gallon by manufacturer, 
time to learn a task by mode of training, frequency of 
appearance of a given type of artifact by geographic 
location, and disposition toward some object or behavior
according to psychological type.  No assumptions are 
made about the distribution of scores on the continuous
variable: they can be normal, multimodal, skewed, etc.  
Two types of analysis are included in this version of 
Random Sampler: Salience Analysis and Effect Size 
Analysis.  In Salience Analysis, the empirically 
obtained values for mean, median, mode and variance for 
a subgroup are compared with values one obtains by 
taking a large number of random samples from the total 
pool of scores.  The probability value obtained 
indicates the degree to which the empirical sample 
stands out from the total group.  In Effect Size 
Analysis, a case is selected randomly from each category
of cases. After a large number of such samples are drawn,
the probability that a given category is greater than, 
less than or equal to each other category is computed as
are the probabilities of the target category having the 
highest or lowest score.   

In addition to the procedures mentioned above, Random 
Sampler includes some auxillary procedures that might be
useful in order to decide how to carry out the analyses.
These include descriptive statistics, cross-tabulations,
breakdown, correlation and regression analysis.  
Procedures are also included to generate random sets of 
integers and to take random samples from a set of data 
in a database with specifiable sample sizes and numbers 
of samples.  Graphing capabilities are provided for 
frequencies, means, medians, modes, variances and 
scatter plots.

Random Sampler will accomodate dBase, Paradox and Ascii
files.  A routine is included to easily convert Ascii 
files to Paradox format.  Data for new cases may be 
added to database files and data for existing cases may 
be edited or deleted.  New variables may be added,  and 
existing variables may be deleted and edited in Random 
Sampler.  Residuals from regression analysis can be 
added to the database with a click of the mouse.

Random Sampler is a MDI (Multiple Document Interface) 
application which means that the user may work with more
than one database on the screen at the same time.


Installing Random Sampler

Random Sampler and its installation program are both 
Windows applications, so Windows must be running in 
order to install Random Sampler.  The installation 
program creates directories and copies files from the 
distribution disk to your hard drive.

To install Random Sampler,

1.  Start Windows if it is not running on your computer.

2.  Insert the Random Sampler distribution disk into 
your floppy drive.

3.  Use Program Managers File| Run menu command or File 
Manager to run SETUP.EXE from the distribution disk.  
In Windows 95, click Start and then Run...and use the 
browser to locate SETUP.EXE on your floppy disk.

4.  Follow the instructions presented by the 
installation program.

The installation program copies files as follows:
rs.exe		to	C:\Random (or whatever directory you choose)
rs.ico		to		    "	
readme.txt	to                  "			
cars.txt	to 	C:\Random\Data 
carstx~1.sch	to		    "
stocks.db	to		    "
talent.dbf	to		    "
bivbx11.dll	to	C:\windows\system
bigauge.vbx	to		    "	
chart2fx.vbx	to		    "
rshelp.hlp	to	C:\windows


Changes from Prior Version

Version 1.52 corrects a bug in the New File routine.  
When attempting to create a new database, the program
would crash.  My thanks to Jonathan Leather for 
pointing this problem out to me.

Version 1.51 corrected a bug in the regression program
of Version 1.5.  When a single predictor variable was
selected, the dependent and independent variables were 
inverted, causing the unstandardized regression coefficient
to be in error.

I am indebted to Mr. James Byrd who tried Random Sampler
on some very large data sets and pointed out a number of
bugs and limitations.  As a result, new data structures 
and procedures were used in computing correlation and 
regression values, making execution much faster.  A 
change in the Salience Analysis procedure also makes for
more efficient performance.

The correlation table was divided into two triangular 
matracies with the upper matrix containing r values and
the lower containing Ns and significance levels.

Code was added to allow running Random Sampler "in the 
background" rather than having exclusive control of the
computer during procedures that are likely to take 
significant amounts of time.  Click on the minimize icon
and you can do something else while awaiting the results.

Printing was completely revised using a Delphi component
developed by W. Murto.  Thanks, Bill!  

The number of intervals to be used in dividing values 
between minimum and maximum can now be specified for
continuous variable data in the Salience Analysis, 
Descriptive Statistics and Graphing procedures.  This 
will make a difference only with respect to calculation 
of medians and modes.

A problem was fixed in the input error checking routine
of the Calculate Values equation dialog which caused 
certain values to be rejected.  

A procedure to draw case samples (including all the 
data associated with cases in the sample) was added to
the Samples page.

