logistic regression with PLINK

Question

[charlie.gaoc]

Guys, I'm running logistic regressions with plink. From the current output, I can tell the total number of individuals (case+control) from the NMISS column. My questions is, is there a way to get the number of cases and controls for each SNP as well? For example, with SNP123456, I currently know the sample size is 5,000. What's the easiest way to tell how many cases and how many controls within the 5,000 people? Greatly appreciate your help! Charlie

P.S. we dropped some genotype calls due to bad quality and we don't plan to impute the missing calls. Therefore, each SNP may have a different number of cases and controls.

[charlie.gaoc]

Thanks. --freq gives nice information based on allelic information. It will be nice for me to get the numbers of individuals based on disease status/phenotype. For example, I need to know how many T2D cases and T2D controls for each SNP analyzed.

[GenoMax]

Please use ADD COMMENT/ADD REPLY when responding to existing posts to keep threads logically organized.

Answer 1

[chrchang523]

With plink 1.9, "--freq case-control" stratifies the allele frequency report by disease status.

With plink 2.0, you can embed case/control allele counts directly in the logistic regression report by adding "cols=+totallelecc" after --logistic/--glm. (Note that since these are allele counts, not sample counts, the numbers will generally be twice as large.)

[charlie.gaoc]

Thank you. I guess it's the best option I have for now. Greatly appreciate it!

Answer 2

[Maxime Lamontagne]

Maybe the "--freq" command will give you this info.