Hi Everyone,
I'm involved in planning a study which is going to test some
hypotheses with quite a lot of potential outcomes - several
drugs/substances choice - alcohol, cigarettes, marijuana, other, with
several possible measures - age at first use, freq*quantity,
dependence, age at dependence, etc.
So we plan to use the Benjamini-Hochberg false discovery rate (FDR)
p-value correction. But then what value does one enter for the power?
I did a bit of googling, and found a paper in last year's Annual
Review of Psychology, by Maxwell, et al, which says: "Although
controlling the FDR tends to yield more power than controlling the
familywise error rate, as of yet no formal sample size planning
procedures currently exist for the FDR."
Oh.
OK, so are there any informal ones? I guess that we could either go
with 0.05, or go with the minimum that the study could give us, or
have the average, or something else. Any thoughts / opinions / ideas
on what's best? Or to help me decide what's best?
Thanks,
Jeremy
--
Jeremy Miles
Psychology Research Methods Wiki: www.researchmethodsinpsychology.com
|