[BioC] problem to compute copy number with crlmm

[Robert Scharpf]

> Hi,
> I'm trying to compute the copy number for the Illumina 660W-Quad  
> bead array
> with the package crlmm. But for some chromosome it doesn't work and  
> I get the
> following message:
>
>> computeCopynumber(crlmmSetList)
> 'batch' missing.  Assuming all samples in the CrlmmSetList object were
> processed together in the same batch.
> Fitting model for copy number estimation...
> Using 50 df for inverse chi squares.
> Sufficient statistics
> .
> Estimating coefficients
> .Called from: nuphiAllele(p = p, allele = "A", Ystar = YA, W = wA,  
> envir =
> envir)
> Browse[1]>
>
> What shall I do? When I press enter I get:
>
> Error in nuphiAllele(p = p, allele = "A", Ystar = YA, W = wA, envir  
> = envir) :

Benilton has removed the browser from the devel and release versions.

crlmm uses the MAD as a robust estimate of the within-genotype spread  
of intensities on the intensity-scale.  These estimates are used as  
weights in a regression model.   What has most likely happened is that  
some of these estimates were zero and the inverse variances (the  
weights) are INF.  We could replace the zeros with some constant to  
prevent the error and replace the stop() with a warning() message, but  
usually the error is a side effect of something more serious --  I  
would like to understand why the zero's occur in the first place.  I  
will try to reproduce the error, but I need to know

- how many samples you are processing with crlmm

- your sessionInfo()

Thanks,

Rob


>  Inf values in W or V
>> traceback()
> 5: stop("Inf values in W or V")
> 4: nuphiAllele(p = p, allele = "A", Ystar = YA, W = wA, envir = envir)
> 3: coefs(plateIndex = p, conf = conf[, plate == uplate[p]], envir =  
> envir,
>       CONF.THR = CONF.THR, MIN.OBS = MIN.OBS)
> 2: .computeCopynumber(chrom = CHR, A = A(ABset), B = B(ABset), calls =
> calls(snpset),
>       conf = confs(snpset), NP = A(NPset), plate = batch, envir =  
> envir,
>       SNR = ABset$SNR, bias.adj = FALSE, SNRmin = SNRmin, cdfName =  
> cdfName,
>       ...)
> 1: computeCopynumber(crlmmSetList)
>
>
> What could be the problem? I'm glad for any help.
>
> Thanks in advance
> S?ren