Hi everyone,

In my data, 10 patients had RNASeq performed before and after treatment. 7 of the patients responded to treatment, while 3 of the patients did not respond to treatment.

I want to look at genes that change during treatment only in responders v.s. nonresponders, and vice versa.

Would ~patient_id + timepoint + response be an appropriate design for this data?

Thanks.

Response and patient are confounded. You'll need some trickery as described in the vignette to include all that in the design. Also, you might want to add the response:time point interaction term too. But it depends exactly what contrasts you want.