[BioC] eBayes warning: zero sample variances

James W. MacDonald jmacdon at med.umich.edu
Wed May 2 19:42:49 CEST 2007


Hi Matt,

Matthew Scholz wrote:
> Hi,
> 
> I'm using R2.3.1 and Bioconductor 1.8 (I think, but am not sure how to
> check). I'm getting warning message that I have not found described in the
> mailing list archives or in the help for eBayes and was hoping someone could
> help.
> 
> Here's the relevant background output, minus a filter function called
> filters that I defined but have left out here for brevity:
> -------------------------------------------------------------
> 
>>targetsA <- readTargets("targetsA.txt")
> 
> 
>>targetsA
> 
>      SlideNumber FileName  Cy3  Cy5
> A296           1 A296.gpr DH1C DH1D
> A297           2 A297.gpr DH1C DH1D
> A298           3 A298.gpr DH1D DH1C
> A299           4 A299.gpr DH1D DH1C
> 
> 
>>RGAn= read.maimages(targetsA$FileName, source="genepix", wt.fun=filters,
> 
> other.columns=c("F635 SD",
> + "B635 SD", "F532 SD", "B532 SD", "F Pixels", "B Pixels", "B532 Mean","B635
> Mean","% > B635+2SD", "% > B532+2SD"))
> Read A296.gpr
> Read A297.gpr
> Read A298.gpr
> Read A299.gpr
> 
>>RGAn <- normalizeBetweenArrays(RGAn, method="scale")
> 
> 
>>designmatrix <- modelMatrix(targetsA, ref="DH1C")
> 
> Found unique target names:
>  DH1C DH1D
> 
> 
>>designfitA <- lmFit(RGAn, design=designmatrix, method="ls")
> 
> 
>>contrastfitA <- eBayes(designfitA)
> 
> 
> Warning message:
> Zero sample variances detected, have been offset
> ------------------------------------------------------------------------
> Do I need to worry about this message? Things downstream seem to be working,
> but appearances are sometimes deceptive.

No. This just means that for at least one gene the log ratio is 
identical for all samples. Since this will give a zero variance (which 
will end up in the denominator of your statistic and could possibly 
result in an infinite value for your test statistic) it has been offset 
to a small value to prevent that possibility.

Best,

Jim


> 
> Thanks,
> 
> Matt
> 
> 


-- 
James W. MacDonald, M.S.
Biostatistician
Affymetrix and cDNA Microarray Core
University of Michigan Cancer Center
1500 E. Medical Center Drive
7410 CCGC
Ann Arbor MI 48109
734-647-5623


**********************************************************
Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues.



More information about the Bioconductor mailing list