Power Analysis In Multilevel Models

Question

[Andrews]

Hi

I am performing a series of multilevel models that are investigating an association between gene-environment interactions and cogntive change over time.

I have ran all my models and I am now interested in performing a post hoc power analysis to determine what effect size I am able to detect at 80% power with a sample size of ~1700.

Does anyone know of any resources that are avalible which are able to do this?

Thanks

Answer 1

[Devon Ryan]

G*Power proves both surprisingly easy to use and also surprisingly powerful time and again for me. I use that to perform power calculations for pretty much all of our animal usage applications (we do a lot of hippocampal-dependent behavioral tests while treating the animals in a wide variety of manners), since it can handle linear models without any fuss (it also has a number of quite handy tutorials).

BTW, be careful with post-hoc power analyses. I consider the wonderfully written paper by Hoenig & Heisey to be required reading, if you've not done so.