[BioC] Limma: Array weights question

[Gordon K Smyth]

Dear Fraser,

The error means that one of internal design matrices in arrayWeights (the 
one used to model the variances) has gone singular.  It is hard to 
diagnose the problem exactly, but I suggest you try a smaller design 
matrix for which all the columns are estimable.

Although this is a slightly different issue, arrayWeights generally works 
best when the number of coefficients is much smaller than the number of 
arrays.  You might even use a smaller design matrix for arrayWeights than 
you use for your lmFit.

Best wishes
Gordon


> Date: Fri, 6 Feb 2009 09:41:47 -0500
> From: "Sim, Fraser" <Fraser_Sim at URMC.Rochester.edu>
> Subject: [BioC] Limma: Array weights question
> To: <bioconductor at stat.math.ethz.ch>
> Message-ID:
> 	<82377DC24E19614291D0B5A4A89DAD74DBAB95 at e2k3ms5.urmc-sh.rochester.edu>
> Content-Type: text/plain;	charset="us-ascii"
>
> Hi,
>
> I am trying to use arrayWeights on a design as follows:
>
>> design = model.matrix(~-1+phenotype+SID)
>
> Where SID refers to matched samples and phenotype as a cell phenotype
> obtained from those samples.
>
> I run array weights:
>
>> arrayw <- arrayWeights(eset, design = design, trace = TRUE)
> Coefficients not estimable: SIDC4 SIDP5 SIDSG7
> Error in chol.default(Zinfo) :
>  the leading minor of order 48 is not positive definite
>
> What does this error mean?
>
> Here is a summary of the samples. The phenotypes are labeled a through
> k.
>
>> colSums(design)
>     a      b      c      d      e      f      g      h      i      j
> k
>     3      4      4      4      4      5      5      5      5      5
> 5
> SID129 SID139  SID92  SIDC1  SIDC2  SIDC3  SIDC4  SIDP1  SIDP2  SIDP3
> SIDP4
>     4      3      4      1      1      1      1      2      2      2
> 2
> SIDP5 SIDSG3 SIDSG4 SIDSG5 SIDSG6 SIDSG7
>     2      4      4      4      4      4
>
>
> Thanks,
> Fraser
>
>> sessionInfo()
> R version 2.8.1 (2008-12-22)
> i386-pc-mingw32
>
> locale:
> LC_COLLATE=English_United States.1252;LC_CTYPE=English_United
> States.1252;LC_MONETARY=English_United
> States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252
>
> attached base packages:
> [1] tools     stats     graphics  grDevices utils     datasets  methods
>
> [8] base
>
> other attached packages:
> [1] limma_2.16.4         RODBC_1.2-4          hgu133plus2cdf_2.3.0
> [4] affy_1.20.2          Biobase_2.2.1        rcom_2.0-4
> [7] rscproxy_1.0-12
>
> loaded via a namespace (and not attached):
> [1] affyio_1.10.1        preprocessCore_1.4.0