What's your power to detect a recessive effect with an odds ratio of 1.2 for a disease with 4.2% prevalence using 1200 cases and 2900 controls? What if the allele is rare? Is it worth it, in terms of power gain, to genotype 1000 more individuals? How small of an effect can you detect with 80% power using the data you have? These questions and others can be answered by power and sample size calculations. While any respectable statistical computing software can do these, it's often far simpler and faster to use "canned" software for power analysis. While there are tons of these on the web, here are a few that I or others around here commonly use.
CaTS Power: This software from the Abecasis lab at Michigan couldn't be simpler to use. Nice GUI with clicky-boxes and sliders, made specifically for power calculations in genetic association studies using case-control study design. Has some nice extra features related to multi-stage designs, or designing penetrance tables for simulation.
PS Power: Made here at Vanderbilt by Bill Dupont in Biostatistics. Not as pretty as CaTS, but much more flexible, allowing for studies with dichotomous, continuous, or survival response measures, as well as matched designs.
G*Power: Probably the most flexible of the three mentioned here, allowing for more specific questions to be addressed, but at the cost of a slightly steeper learning curve.
Getting Genetics Done by Stephen Turner is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported License.
I like this title! -EPT
ReplyDeleteR can do most any power calculation you need.
ReplyDeleteHi Stephen. I have basic question. In my linear regression analysis PLINK reported beta as effect. Can I use beta directly for power calculations for quantitative trait. If not then what should be the right way to do it?
ReplyDeleteIt's a place to start but there are many other considerations: allele frequency, how many SNPs you're testing, inheritance model, etc. Have a look at this paper and others on the topic to fully appreciate what goes into power calculations for genetic association studies.
ReplyDelete