This doesn't really have much to do with minfi, but instead has to do with how you specify/interpret linear models in R. In other words, minfi is doing special things to smooth data across CpG sites, but when you fit the model, you are just fitting a linear model on data, using R.

While you can often get answers to explicit modeling questions on this forum, in general the expectation (for anybody who is analyzing data, really) is that the analyst knows what he/she is doing. Unfortunately, the aspects of linear modeling, plus how models are implemented in R, is something that is WAY beyond the scope of a support forum. Linear modeling alone is like four or five of the courses I took to get my master's, and that didn't include figuring out how to do these things in R (back when I was in school, we used this nasty thing called SAS, which makes me shudder to even think about...).

Anyway, there is no substitute for knowing what you are doing, and you aren't going to get that here. Good resources are the Users Guides for limma and edgeR, the DESeq/DESeq2 vignettes, which have lots of examples. Julian Faraway's book is also a really good resource. Plus I am sure you can dig up other things using the googles.

There's also this book, hot off of the presses (well, still on the presses, but you can get a sneak peek):

https://leanpub.com/regmods

Jim, Steve, thanks for your constructive advice, I'll admit, I've been teaching myself a lot of linear model design / interpretation to try and really understand the inner workings on Limma, so those resources are great.

Based on my OP, I don't think I phrased my question very well, I guess what I was looking for, was to see if there was a way that Bumphunter dealt with contrast matrices, but in hindsight, that was just the wrong way to approach things.

I'm in the same boat as you: much of my linear modeling chops was pretty much self taught. To be honest, I've learned quite a lot by just lurking on this mailing list/forum over the years, so I need to periodically thank the good folks like Jim, Gordon, Simon, Michael, Ryan (and more recently) Aaron for taking the extra time to provide some applied linear modeling schooling here ;-)

Couldn't agree more!