Since you have lots of possible effects, you need to decide what effect sizes for each that you want to be able to detect, and how many of these that you want to detect. G*Power is more designed when there is a single effect of interest. Once you decide these, then I recommend using simulation to estimate power.

Since you have lots of possible effects, you need to decide what effect sizes for each that you want to be able to detect, and how many of these that you want to detect. G*Power is more designed when there is a single effect of interest. Once you decide these, then I recommend using simulation to estimate power.

Hello Tabir,

Daniel Wright is correct; you'll need to determine: (a) which effects, out of the four main effects possible, the six two-way interactions possible, the four three-way interactions possible, and the possible four-way interaction, you really wish to evaluate; (b) establish the target effect size (in G*Power, you'll be using Cohen's f or f-squared as ES metric) for each effect to be evaluated--what is the smallest, non-ignorable effect that you'd like your study to be able to detect, if it truly exists; then (c) run G*Power or whatever software you select (and, yes, simulation is indeed a viable option); then (d) base sample size on the largest resultant minimum N computation that results.

Good luck with your work.

And to add to David Morse 's list, decide how you will adjust the multiple p values.