This may be a lame question, but I got stuck and can't get my head around it. I am running a gene expression analysis, comparing $\sim 10,000$ genes between two groups, $n=6$ samples per group. My pipeline goes like this:

  1. I check whether the two groups have equal variances using Levene's test for each gene.
  2. For those genes that do have equal variances, I run a moderated t-test (as in limma), as it assumes equal variances.
  3. For the rest of the genes I run Mann-Whitney U.

(Before you tell me about multiple test correction, I do calculate FDR later on by permuting the population labels on samples and re-running the above tests).

Now, I want to calculate effect size for all the genes. I am using Cohen's d with pooled variance (see Wikipedia), but I wonder whether I can do it for both groups of genes (ie ones with and without equal variance between groups). I think I can't, but it's a guess really (I think I can only run it for the equal variance genes). If I can't, is there a more appropriate way to do this?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.