I want to replicate the factor analysis process of one published paper,
This process is mainly to calculate the unifying gene expression matrix from 3 different platforms but with same patient samples.
However, I tried many times but failed to get the right answer, the whole raw data and processed data are offered here
First I log transformed the "Broad202.txt" since it is not done yet, then I tried several method: 1) cbind the expression data for each gene across 3 platforms to build the matrix to package "factanal" 2) first scale the expression matrix to mean 0 and sd 1, then do 1) 3) first calculate the relative expression data for each gene, using (x-mean(x))/sd(x)
But none of the methods above can get the results of unified gene matrix
And actually I think the 3) is one step in factanal, so it should be not right.
If you have tried this kind of process before, could you give me some hints?
