remove B for irwsva.build

[yufree]

Dear Prof. Leek ( @jtleek ) ,

I am working on a metabolomics project and read your excellent papers on SVA. I think SVA could be applied to metabolomics studies to remove batch effects.

I found parameter B in both num.sv and irwsva.build function. They are different (one for permutation and one for iteratively re-weighted) while the same name might make others confusion in sva function.

Then I found B in irwsva.build could be substituted by a while loop. I constructed another weighted matrix for the mod by:

dats2 <- dat*pprob.b*(1-pprob.gam)
dats2 <- dats2 - rowMeans(dats2)

Then I use a while loop based on this matrix to test if some surrogate variables are still there. If the numbers is other than 0, the reweighting process will keep running.

while(num.sv(dats2,mod,method="be") != 0)

By this way, the iteratively re-weighted feature could be retained and use less parameters for users.

I wonder if this method could be merged into the devel version of SVA?

Thanks and Sincerely,

Miao