I don't feel very comfortable with stats (still using simple programmes like SPSS rather than R!), and have a question about which is the best test to use with the following design:
I have performed an experiment with n = 4 per group (three groups: drug A+ drug B treated, drug A+ vehicle treated, vehicle only).
I've then extracted chromatin (= nuclear protein-DNA complexes) from these samples (12 in total), used antibodies to isolate bits of DNA and checked whether particular genes are more highly represented (= enriched) over others.
What I want to test primarily is: 1) Is there a difference in gene enrichment between vehicle only and drugA+vehicle treated groups, and if so, which genes? 2) Is there a difference in gene enrichment between drugA+drugB treated and drugA + vehicle treated groups, and if so, which genes?
I have tested 5 genes for each sample, and to complicated matters further I have tested two antibody batches (they are meant to contain the same thing, but as the lot differs they might not, and indeed they don't quite give me the same results).
Initially, I was going to run a simple multivariate ANOVA with treatment group as the fixed factor, where I just checked for every paramater separately gene 1,antibody batch 1 gene 1, antibody batch 2 gene 2, antibody batch 1 gene 2, antibody batch 2 etc
However, I wonder whether I lose power like this, because obviously I expect the result for the two antibody batches to be somewhat related. I also to some extent would expect the effect on the various genes to be somewhat related within a sample, as these genes all belong to a particular biological network.
However, I suppose it is wrong to use a repeated measures test (e.g. with treatment as the fixed between-subjects factor, and gene and antibody as within-subjects factors?)
Would you recommend some kind of mixed model instead?
I hope this is clear enough...it is rather more complicated to explain than I thought.