Hi, I have 2-factors (treatment and task) with repeated measures design. Each subject has to preform 2 different tasks before and after treatment. We have 5 subjects, so it is 20 arrays in total. For each probeset, I have tried with aov.res <- aov( expr[1,] ~ (treat * task) + Error( subject / (treat * task) ), rma_norm ) Now, I would like to translate the above model into LIMMA. I was bale to do run LIMMA without the error term. How can I include the error term in LIMMA? I greatly appreciate your suggestions. Thanks. Best regards, Sharon [[alternative HTML version deleted]]