[BioC] lmFit, weights and A

[Marcus Davy]

Additionally to Seans comments, the row mean of the A matrix is all that is calculated in lmFit as

  if (length(object at maA)) 
            fit$Amean <- rowMeans(unwrapdups(object at maA, ndups = ndups, 
                spacing = spacing), na.rm = TRUE)
    }

which are not weighted, and the Amean is reported in topTable (if it in the list). In plotMA(fit) you will have 
some small differences in the Amean's xcoordinate compared to using a weighted mean to calculate A values
when highlighting genes predicted to be differentially expressing.

Marcus


>>> Sean Davis <sdavis2 at mail.nih.gov> 18/05/2005 11:20:05 p.m. >>>

On May 18, 2005, at 4:18 PM, lepalmer at notes.cc.sunysb.edu wrote:

> I was playing around with weights and examining the results.   I 
> changed
> the wieghts prior to doing duplicateCorrelation,lmFit,eBayes and 
> topTable
> in limma.  Changing all weights to 1 (were previously they were 1 or 0)
> had  a dramatic affect on M values, but did nothing to A.    Is this
> intended?  I would think that if you flagged something, you would not 
> want
> the results included in the calculations for A.

The A values do not enter calculations of duplicateCorrelation, lmFit, 
eBayes, or topTable, as far as I know.  They are reported in topTable, 
yes, but A values don't affect which genes show up as differentially 
expressed.  Is there a specific reason you want to know the "weighted" 
A-values--just curious?

Sean

_______________________________________________
Bioconductor mailing list
Bioconductor at stat.math.ethz.ch 
https://stat.ethz.ch/mailman/listinfo/bioconductor

______________________________________________________

The contents of this e-mail are privileged and/or confidenti...{{dropped}}