Hi all I have a design that I'm wondering if someone can help me double check.  Basically I have he following groups. 

Adult.autism 
Peds.autism
Adult.control
Peds.control 

I want to do the following comparisons

* compare ped control vs. ped autism 
* compare adult control vs adult autism
* ped autism vs adult autism 
* ALL control vs ALL autism regardless of age.

Here are my codes and design.  Can someone help confirm if this is correct?  especially the last all group. thanks.  

group <- factor (key)
design <- model.matrix(~0+group )
colnames(design) = gsub ( "group","", colnames(design))
fmain  <- lmFit(voom.data, design )

contrast.matrix <- makeContrasts(p = ped.control - ped.autism, 

                                 a = adult.control - adult.autism,

                                 pVSa = ped.autism - adult.autism,

                                 ALL = (adult.autism + ped.autism) - (adult.control + ped.control ), 

                                   levels=design

                                   )

  

fmain.r <- contrasts.fit(fmain, contrast.matrix)

fmain.r <- eBayes(fmain.r)

result.p <- topTable(fmain.r, coef="p", n=Inf )
result.a <- topTable(fmain.r, coef="a", n=Inf )
result.pVSa <- topTable(fmain.r, coef="pVSa", n=Inf )
result.ALL <- topTable(fmain.r, coef="ALL", n=Inf ) 

I don't know what lmfmain or contrasts.fmain are. Aside from that, your design matrix looks fine. For your ALL contrast, you should divide by two as you're taking the average of the with/without autism groups:

ALL = (adult.autism + ped.autism)/2 - (adult.control + ped.control )/2

Otherwise your log-fold changes will be twice as large as they should be.