Post hoc power analysis

#1
Hi,

I need to do a power analysis in G*power for my study. My research is part of a larger study for which the data is already collected, so my sample size is predetermined for me. I don't have access to the actual data yet. (note: I will be looking at a patient group and a control group over 3 different time points. The patient group have already completed treatment prior to the first time point and so there is no interventions during the study, it is just observational of each group's self reported symptoms over time). Where do I get the d value from in order to conduct this post hoc power analysis? Also which settings should I be selecting in g*power? Does it depend on which analysis I am going to run? I am thinking a repeated measures ANOVA might be best but also happy for advise on this.

Thanks in advance for any guidance anyone can provide
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
I wouldn't call it post hoc, since you haven't ran the study analyses yet. Post hoc power analyses are a major faux pas, since most people do them after finding significant results.
 
#3
I wouldn't call it post hoc, since you haven't ran the study analyses yet. Post hoc power analyses are a major faux pas, since most people do them after finding significant results.
Hello, thanks for your response. The study has been run and the data are sitting with the larger institution, I'm just personally not allowed to view them for a few more weeks. Do you have any insight how I can assess the power or expected effect size for my study? I am expected to do this.
 

Karabiner

TS Contributor
#4
I guess that you should just perform an a priori power analysis. The analysis seemingly
is a mixed anaylsis of variance (with a between-subject factor "group" and a within-
subject factor "time"). Whis effect size is reasonable to assume, or desireable to achieve,
depends on the area of study and on the study goals. Maybe g*power has some device
built in which allows you to calculate an effect size statistic, or you just use the "medium"
effect size.

With kind regards

Karabiner