Maren Kandulla wrote:
I am doing an evaluation about different
curricula in medical education. I have four different treatment groups
(students) to compare and longitudinal data for 7 different measuring times.
Since we have small sample size I would like to calculate the power for our
tests.
Does anybody know how to calculate power (or samplesize) for a four-sample
comparison with repeated measures?
I found that:
sampsi mean1 mean2, sd1() sd2() method(change) pre(x) Post(y) r1(z) n1()
n2()
works for two-sample comparison with repeated measures but not for a
four-sample trial.
The ADOs: simpower and ssizebi are brilliant for a four-sample comparison,
but do not offer a method for repeated measures.
Maria's immediate problem can be addressed by using
sampsi, noting the adjusted standard deviation,
and then using this in simpower. (As power is likely to be limited,
I would advise method(ancova) rather than the less powerful
method(change) in both power calculations and eventual analysis.)
However, I am not sure whether a series of pairwise calculations
might not be be better. After all, if the ANOVA it is significant,
you need the results from the pairwise comparisons before you
can say anything useful, so you really need the pairwise
comparisons to be adequately powered.
*****************************************************************
Maren Kandulla
Wissenschaftliche Mitarbeiterin / Evaluation
Modellstudiengang Medizin
Universit�tsklinikum Hamburg-Eppendorf
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
