I have 4 dose groups, technical duplicates, and pre- and post-treatment data. I'm interested in detecting a linear trend across dose groups, adjusting for pre-treatment variability. I'm not certain how to set this up. Data is like this:
patients <- data.frame(Patient = rep(c(1,2,3,4,5,6,7,8), each=4), Treatment = factor(rep(c("Placebo", "Dose1", "Dose2", "Dose3"), each=4), levels = c("Placebo", "Dose1", "Dose2", "Dose3"), ordered=TRUE), Visit = factor(rep(c("Baseline","Baseline","Visit1","Visit1"), 8)))
Is this sufficient?
design <- model.matrix(~ Visit + Treatment, data=eSet)
corfit <- duplicateCorrelation(eSet,design,block=patients$patient)
fit <- lmFit(eSet, design,block=patients$patient,correlation=corfit$consensus,
fit2 <- eBayes(fit)
topTable(fit = fit2, coef = "Treatment.L" )
Or does there need to be a Treatment * Visit interaction, then look at the Treatment.L * Visit1 effect?