送交者: xj 于 2005-8-21, 21:18:01:
回答: Thank you. Copied and pasted. 由 cornbug 于 2005-8-21, 20:04:20:
Because I am working in the social science field, I don't do much about multivariate ANOVA although I know quite a lot about repeated stuff, or panel data.
In fact, Mango is also able to help you if I didn't speak out. She must have learned panel data analysis too, except she may not be familiar about MANOVA terminology.
Calculating sample size is a trade off between resource and statistics. But be serious about sample size. Without sufficient large sample size, you can't obtain significant results, thus you can't publish any results even they may be suggestive.
Ironically, you never feel you have gotten enough data. the larger sample size, the more subgroup analyses and complicated analyses you will propose. In a sense, sample size is never too large.
In addition, there must be someone to do the experiment to collect the data. It is better you do it rather than somebody else.
In social science, we don't do much about sample size stuff. If the data are there, then everything is there. Why bother? :-)
The practice of statistics is an art, not a hard science.
Also thanks for the update on gpower, it surprised me that there is quite something in gpower. I really appreciate the formula it gives. It saves my time to find them.
Whenever I consult something, I learn more than I have offered.