Not sure why you had trouble with InerSTAT-a. The inputs are clearly
labeled. You do not input the dataset, you input summary statistics
(treatment means, standard deviations, and sample sizes).
However, InerSTAT-a v1.3 calculates critical values for Tukey's HSD from
only 3 terms of an asymptotic expansion. Consequently they are
inaccurate for small degrees of freedom. InerSTAT-a v1.3 results should
be reliable for df=10.
The table at
http://web.umr.edu/~psyworld/virtual...icaltable.html
should be accurate to all figures given, since it is an accurate subset
of Table 29 from the 3rd edition of "Biometrika Tables for
Statisticians" (BTKS3).
I do not know how accurate the p-values calculated by prtrng from
http://lib.stat.cmu.edu/apstat/190
are, but they seem consistent with BTKS3. Critical values calculated by
qtrng are less accurate than numerically inverting prtrng p-values.
Jerry
Jerry W. Lewis wrote:
Sorry, a Google search shows that your description is apparently specific in
the psychology literature to what the statistics literature would call
Tukey's HSD multiple comparison procedure.
The calculations are straightforward
http://davidmlane.com/hyperstat/B95118.html
http://web.umr.edu/~psyworld/tukeyssteps.htm
except for determining p-values or critical values. For that, you could
translate
http://lib.stat.cmu.edu/apstat/190
from Fortran into VBA if you do not care to use a table.
Jerry
"Ross" wrote:
excel does not have such a built-in function so i tried analyse-it and
INERST13.XLS from web. both don't have any clear instructions how the
dataset should be input. is anybody knowing how to best perform such a test
in Excel? thx in advance!!