Sample size calculation

[sabah1023]

Hi,
I have a 2X2X2 design and I want to get medium effect. This is a 3 way ANOVA and I am not sure how to calculate the sample size in G*Power. Not sure what "number of measurements" stands for in G*power. Can anyone please help me calculate the sample size? Thanks!!!

[Bruce Weaver]

Your ANOVA will yield 7 F-tests (3 main effects, 3 two-way interactions, and one three-way interaction).  Which one (or ones) of these do you want to have adequate power for?  

Also, not everyone recommends using standardized effect sizes as the basis for sample size estimation.  For example, http://homepage.stat.uiowa.edu/~rlenth/Power/2badHabits.pdf.  

HTH.

sabah1023 wrote

Hi,
I have a 2X2X2 design and I want to get medium effect. This is a 3 way ANOVA and I am not sure how to calculate the sample size in G*Power. Not sure what "number of measurements" stands for in G*power. Can anyone please help me calculate the sample size? Thanks!!!

[Art Kendall]

What is your dependent variable? How is is it measured? How many legitimate values does it have?

Are you primarily interested in an interaction effect, e.g, a difference in changes for  experimental vs control (or other comparison group)?

Do you have random assignment  to treatment?

Are your factors between or within subjects?

On the metric of your DV(s) what size difference (change?) would make a theory/policy/practice difference?  This asks about meaningful  differences not just whether or not observed difference are attributable to chance variation.  In psychology this last is known as the "So what test" which needs to be applied if the differences are found to be "statistically significant".