Hello,

I have a question in relation to limma and how to extract the random intercepts results after running a duplicated correlation. To summarize, I have DNA methylation data from a repeated intervention that was inter-spaced by one year of wash out. What I would like to see is if DNA methylation markers change similarly even after one year of wash out when the same intervention is applied again. My code is as follows:

design=model.matrix(~intervention+Age+Batch,data=pheno) 

library(limma)

corfit <- duplicateCorrelation(DMFS,design,block=pheno$ID)

fit = lmFit(DMFS, design,maxit=1000,block=pheno$ID,correlation=corfit$consensus) 

fit2 <- eBayes(fit)

Where the outcome is delta DNA methylation (Delta M-values), intervention is 1 or 2. Then n the duplicateCorrelation I have blocked by participants ID (random intercept). However I can seem to find a way to extract results from this random intercept. I would like to see if any CpGs (probes) change similarly in both interventions, at the individual level. Thus I need estimate, SD and p-values for the random intercept. Can anyone help please?

First, I need to point out that duplicateCorrelation fits a mixed linear model random effect. A "random intercept" is something different that is not needed in the limma context.

Sorry, but you are asking for things that don't exist. Random effects don't produce p-values. It just isn't mathematically meaningful to include participant as a random effect but also try to test for differential methylation for individual participants. Testing for results at the individual level is also not a normal thing to do from a scientific point of view.