Dear Jing, thanks for your message, i'm including the BioC mailing list in the cc in order to keep an archived copy of this sort of communications about the BioC package qpgraph so that other BioC users encountering this problem in the future can benefit from our discussion, and so i'd appreciate if you also keep this mailing list address in your reply. the error below indicates some problem with inverting a matrix of size q +2 which because the output indicates that q=1 was being used we're talking about a matrix of size 3 and thus it's obvious that the problem is not about specifying a wrong q value but rather I suspect it may have something to do with the type of data that you're providing as input. you mention that the data consists of clinical phenotypes and metabolomic profiles. the qpgraph method assumes that the data is a multivariate normal sample, if your clinical phenotypes are some kind of discrete data, that may be the reason. let me know if all your data consists of continuous variables because then it might be a bug in the software. best regards, robert On Wed, 2010-04-14 at 15:32 +0300, Tang Jing wrote: > Dear Dr Castelo, > I am currently using the qpgraph package for analyzing a large data > with 122 subjects and 1507 variables. But it returned an error messge > when calling the qpAvgNrr function as below: > ------------------------------------------------------------- > q=1 > 0Error in qpgraph:::.qpFastNrr(S, N, q, nTests, alpha, > pairup.i.noint, : > Lapack routine dgesv: system is exactly singular > -------------------------------------------------------------------- ------------------------------------------------------ > Could you please shed some light on the reasons? These variables are > clinical phenotypes and metabolomic profiles. > > Thanks! > > Jing Tang > Researcher > VTT Finland > > >