Hi everyone, I'm looking at the example of a complex experimental design on the most recent (April 2015) on page 36
The example states "To find genes that respond to the hormone in any disease group:"
> lrt <- glmLRT(fit, coef=10:12)
> topTags(lrt)

I'm interested in identifying the genes that respond to hormone in all disease groups. How is the glm constructed to identify genes that respond in all disease groups?

Thank you!

To do that, you'll have to test for an effect in each disease group, i.e., by dropping each coefficient individually. This will give you three sets of DE genes. You can intersect them to find the common set of genes that respond in all of the three disease groups (you might want to select by the sign of the log-fold change first, to ensure that you're picking genes that respond in the same manner in each group). I don't think it's possible to construct a GLM contrast for a composite null hypothesis.