Hi Dear Friends,

The design of my research is a within-subject 4-way anova.
Each of the four independent within-subject variables has two levels.
I need to do a priori power analysis to determine the appropriate sample size of the study.
If power is set at 0.8, and effect size estimated at 0.5 (medium), what would be the appropriate sample size?

Many thanks.

You have not provided enough information to give a number.
I recommend the following free program.

http://www.psycho.uni-duesseldorf.de/ab ... ap/gpower3

I'm not sure (never really used the program myself), but some work/college friends of mine said that NCSS PASS is really good for power and sample problems. But it isn't free...

sasky's reference to G*Power is the way to go. It's both comprehensive & free.

Hi Sasky, Fierce, Statman,

Grateful! Yes the way you showed me is the way to go.
May I ask to have further advice? I feel I am just one step away from the solution.
I have access to both PASS and G*Power.
PASS accepts only 3 IVs (there are 4 in my research).
G*Power appears user-friendly. However, I followed the link, for my design (ANOVA: Repeated Measures, within subject), the last four boxes on the left column are:
"number of groups", "repetitions", "correlation among rep measures", "nonsphericity correction".
To my understanding, nonsphericity is the violation of identically and independently distributed errors. It is inevitable when the repeated measures data are correlated to a large extent. Could I accept the default of G-power (1)? The default for "correlation among rep measures" is 0.5.
Not sure what "number of groups" and "repetitions" mean. On-line help is not yet available for my design. Please advise.

Thanks to you all!