home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
DP Tool Club 25
/
CD_ASCQ_25_1095.iso
/
dos
/
educ
/
kstat41
/
xxks2.lzh
/
KS41.DOC
< prev
next >
Wrap
Text File
|
1995-08-16
|
11KB
|
296 lines
KWIKSTAT 4.1
Supplemental Manual
Statistical Data Analysis
June, 1995
(C)Copyright TexaSoft, 1983-1995
P.O. Box 1169
Cedar Hill, Texas 75106-1169
214-291-2115 Fax:214-291-3400
INSTALLATION
Note: If you install KWIKSTAT 4.1 in the same directory as version 4.0, the
installation of 4.1 will overwrite your old KWIKSTAT program files, but
will not overwrite any new .DBF (database) files you may have created. By
default, KWIKSTAT 4.1 is set up to install in a new directory called KS41.
Once you have installed and tested the new version, and copied all of your
data files, you can delete version 4.0.
Install KWIKSTAT 4.1 using the same procedure described on page 1-3 of the
main manual. Briefly, make A: (or B:) the default drive. Then, enter
INSTALL, and follow the instructions on the screen.
PROGRAM DIFFERENCES
This manual supplement describes the differences and enhancements from
KWIKSTAT version 4.0 to version 4.1. Although there have been a number of
small cosmetic changes throughout the program, only those that have changed
the program usage in a significant way are described here. Other program
differences or enhancements are described in the file LATENEWS.DOC. Please
print this file and read it.
GENERAL PROGRAM ENHANCEMENTS
If you are performing a standard type of analysis, it is often easiest to
use a pre-defined database structure to create a database. New pre-defined
database structures have been added to the list in version 4.1. These
include a number of new structures for graphs as well as for analyses. To
display this list, choose the New Database option from the DATA menu. A
list of the available pre-defined databases will be listed. Scroll through
1
KWIKSTAT 4.1 Supplement
___________________________________________________________________
this list to see what database structures you can select. Of course, you
can always choose the Custom Database option to create a database unique to
your analysis.
SUMMARY STATISTICS
The limit for number of groups in the "Summary Statistics" option in the
Descriptive Statistics option from the ANALYZE menu has been increased to
1000 groups. Of course, if you have that many groups, the calculation will
take a considerable amount of time.
BY-GROUP PLOT
A new plot type has been added to the "By Group" option in the Graphs
option from the ANALYZE menu. The new plot type displays means with error
bars. Mean values are displayed as bars. This type of plot is often seen in
medical and psychological journals. You can optionally display these plots
with means connected with a line.
FREQUENCY ANALYSIS
The Frequency Analysis program is located in the Crosstabulations,
Frequencies and Chi Square option on the ANALYZE menu. In version 4.0, you
had to select one field at a time to perform a frequency analysis. In
version 4.1, you can select up to 30 fields at a time. When you request a
Frequency Analysis, a pick list is displayed. Select as many as thirty
field names from this list. As you select a name, it is displayed on the
screen in the list. Once you finish your selection, an analysis for all
fields will be performed, and displayed.
NEW 2 X 2 TABLE STATISTICS
Two new statistics have been added to the 2 x 2 crosstabulation
option in the Crosstabulation, Frequencies and Chi Square option on
the ANALYZE menu. These are Relative Risk and Odds Ratio. In a 2x2
table, the relative risk is given by the formula
a/(a + b)
RR = ----------
c/(c + d)
where the two by two table is
2
KWIKSTAT 4.1 Supplement
___________________________________________________________________
Factor 1
+ -
---------
+ |a | b|
Factor 2 ---------
- |c | d|
---------
the odds ratio is calculated by
a / b
OR = -----
c / d
For more information on these two statistics, reference any
biostatistics/epidemiology text. One example is Basic Biostatistics in
Medicine and Epidemiology, A. A. Rimm, Appleton-Century-Crofts, 1980.
NEW SAS CODE OUTPUT OPTIONS
A new SAS export feature has been added to the File Utilities option from
the DATA menu. When you choose to create an SDF (Standard Data File) output
file, and choose to output the data definition to a file, you are given a
choice to output the SAS code to create a standard SAS data file or output
the SAS code to do a sample analysis using the exported data. The SAS code
files created may be used unchanged with most versions of SAS. For example,
if you export the data from the EXAMPLE database to SAS code to create a
standard SAS data file, you would create a file called EXAMPLE.TXT that
contained the data and a file called EXAMPLE.SAS that contained SAS code.
If you are using PC SAS, you could then enter the command
SAS EXAMPLE.SAS
from DOS, (i.e., at C:\SAS>) to run SAS and create a SAS data file called
EXAMPLE.SSD. If you are running SAS on a mainframe, you would need to
upload the files EXAMPLE.TXT and EXAMPLE.SAS to your mainframe, then run
SAS on EXAMPLE.SAS. If you are not using PC SAS or SAS for WINDOWS, you
may need to change the LIBNAME and INFILE statements in the code to run on
your platform.
CONVERSION OF A SAS DATA SET TO .DBF (PROFESSIONAL EDITION)
SAS code to convert a SAS data set into a .DBF file that can be used in
KWIKSTAT is contained in a file called SAS2DBF.SAS. This file is the source
code for a SAS program. You must have the SAS program to run this code.
To use the conversion program, you must edit the file called SAS2DBF.SAS to
change the filename for the SAS ".SSD" data set file, and the references to
the directory containing the file. Documentation in the SAS2DBF.SAS file
points out what changes need to be made. It is important to set formats for
3
KWIKSTAT 4.1 Supplement
___________________________________________________________________
any numeric fields with decimal values. Otherwise, these numbers may be
truncated to integers during the conversion.
ANALYSIS OF COVARIANCE (PROFESSIONAL EDITION)
The "Analysis of Covariance" option, located in the Advanced Analysis of
Variance module, is an analysis technique that combines regression and
analysis of variance. Sometimes, when designing a One Way Analysis of
Variance, you may be able to reduce the error term if you have information
from an independent variable that is related to your dependent variable.
For example, see the Custom Regression Example on page 5-32 in the main
manual. In this case there are 3 treatments (the grouping variable) and you
are interested in seeing if sales (Y) during a promotional period is
greater for a particular treatment (color of package). You also have
another variable -- sales during a previous period, using the same
packaging. (This is called the X or concomitant variable). Data for this
analysis is in a file named ANCOVA.DBF which has the following fields:
COLOR - The grouping variable, which is color of package, 1=Red,
2=Green, 3=Blue.
Y - Sales during a promotional period
X - Sales during a previous period
Open the ANCOVA database, choose the "Analysis of Covariance" option, and
select the fields as described above. Part of the results reported are:
ESTIMATED TREATMENT MEAN RESPONSES AT XBAR
Treatment Mean Response N
--------- ------------- ------
1 37.55068 6
2 35.55799 6
3 26.72466 6
Test for group effects
----------------------
F( 2 , 14 ) = 22.47097 p = 0.043
Test for parallel slopes
------------------------
F( 2 , 12 ) = 1.815916 p = 0.410
Because the test for parallel slopes is not significant and
the test for groups is significant, you may conclude that
there is a significant difference in the group elevations.
Multiple comparison tests are appropriate to decide where
the significant differences lie.
4
KWIKSTAT 4.1 Supplement
___________________________________________________________________
Multiple comparisons (Scheffe')
Critical S
Comparison Difference S (0.05)
---------------------------------------------------------------------
Elev 1 - Elev 2 1.992683 1.140 2.735
Elev 1 - Elev 3 10.82602 6.196 2.735 *
Elev 2 - Elev 3 8.833334 5.234 2.735 *
Differences marked with an * are significant at the 0.05 level.
From this output, you can see that elevation 3 (BLUE color) ranks
significantly lower than the other two treatment types. There was no
significant difference found in sales between colors Red (1) and Green (2).
Another way to see this relationship is by displaying the plot. From the
options menu on the plot, you can also choose to display the trend lines
for the regressions by group.
Comparison of the elevations can be considered appropriate since the test
for parallel slopes is not significant (although in the plot the slope of
treatment 3 seems a bit different that the other two.) This graph visually
confirms that group 3 has an overall mean less than the other two groups,
which would mean that Blue was the worst color in terms of sales for this
packaging.
The Scheffé procedure is recommended for the multiple comparison test (see
Neter, et. al.,1990, p. 877), and is automatically performed for this
analysis.
5