[BioC] GlobalAncova

[Hans-Ulrich Klein]

Dear List,

I have a couple of covariates in conjunction with my gene expression 
values. I want to know, whether some of the covariates have an influence 
on the expression values. The GlobalAncova package seems to be able to 
answer this question.

"data" is a matrix (genes x samples) with gene expression values. 
"pData" is a data frame with several covariates (factors as well as 
numeric values). One factor is "Hyb" and describes the day of 
hybridization. I tried the following:

 > ga.Hyb <- GlobalAncova(xx=data, formula.full=~Class+Hyb, 
formula.red=~Class, model.dat=pData, perm=perm, test.genes=NULL);

The call above leads to the error message
"Error in solve.default(t(D) %*% D) : Lapack routine dgesv: system is 
exactly singular".

Is it possible to use a pseudoinverse within GlobalAncova to get a 
parameter estimation in spite of the singularity? I read that 
GlobalAncova uses ordinary least squares like lm() does. lm() can handle 
such situations.


Regards,
Hans-Ulrich