[BioC] Re: normexp errors

[Gordon Smyth]

On Fri, May 20, 2005 2:29 am, M Perez said:
 > Thanks Gordon for all your answers.
 >
 > I am afraid I have more questions, now about "normexp"
 > background correction, I am trying to use "normexp"
 > with Limma 1.8 and I got this error
 >
 >> RG <- backgroundCorrect(RG, method="normexp",
 > offset=100);
 > Corrected array 1
 > Corrected array 2
 > Error in optim(theta, m2loglik.normexp, control =
 > list(trace = trace),  :
 >   Function cannot be evaluated at initial parameters
 >
 > Note that only the two first files are corrected for
 > background the third file seem to have something
 > wrong, but when I try with Limma 1.96 and I got this
 > new error:
 >
 >>RG <- backgroundCorrect(RG, method="normexp",
 > offset=100);
 > Corrected array 1
 > Error in optim(c(beta, log(sigma), log(alpha)),
 > sumloglik, gr = grsumloglik,  :
 >         initial value in 'vmmin' is not finite

We are working on this bug now. In the meantime, try this:

RG2 <- backgroundCorrect(RG,method="subtract")
RG2$Gb <- RG2$Rb <- array(0,dim(RG2))
RG3 <- backgroundCorrect(RG2,method="normexp",offset=100)

In principle, this is the same as what you've asked for, but it seems to 
get around a bug.

Gordon

 > Now just only the first file could be corrected using
 > "normexp", I think this problem show up because of the
 > nature of my files that they are "ScanArray" format
 > files, as I read in other posts "normexp"  works most
 > of the time but not on
 > all data sets but I would like to know if is possible
 > to do something before to try the "half" correction.
 > Any idea what could be the reason of these errors?
 > Thank you in advance.
 > Manuel