From: Alan Salmoni <SalmoniA@Cardiff.ac.uk> - 2002-05-07 15:27:05
I noticed a small bug in the latest release of SalStat (20020505) -
when selecting an anova (or other test with more than 3 conditions),
data from all the used columns were sent to the test, regardless of
what the user selected (ie, if you had 4 columns of data entered, all
4 were used whether you selected 2, 3 or 4). This has been fixed, and
will be available in the next release which should be either the end
of this week or the weekend/next Monday. The anova routines have also
been updated to be a bit faster, and more testing has been done on
There was also a weird bug that affected the output of the Spearmans
and Kendalls correlations when a Kolmogorov-Smirnov test was also
selected, but this has also been sorted out and will be available
next release also.
I have been tested the output with the output of several other
packages (SPSS, Statview, Statistica and Minitab) for a consensus,
and shall present the results very soon for comparison. Apart from a
couple of minor errors (and the Freidman test not working at all!),
everything seems to be on course. For example, t-tests
(paired/unpaired), single factor anova (within and between subjects),
and Pearsons correlation produce results exactly like SPSS with one
difference: The p-value is expressed to 6 decimal places for greater
I am also having second thoughts about putting all the "similar"
tests in the same dialog - any opinions welcome as to whether this is
good or would it be better to have a separate (but consistent) dialog
for each test like existing packages?
Alan James Salmoni
Get latest updates about Open Source Projects, Conferences and News.