Entering edit mode
Dear Matthew,
Your question relates to the limma package. contrasts.fit() includes
all
information, including info on genes with missing values. You can use
topTable() for example to rank genes, including those with missing
values.
However eBayes() does not compute F-statistics for genes with missing
values. I agree that this is a restriction, but I don't have any
plans to
change it, because some methods in limma that use F-statistics expect
to
have full df.
Personally, I like to preprocess microarray data so that missing
values
are not introduced in the first place. That would be one possible
solution.
Best wishes
Gordon
> Message: 1
> Date: Mon, 22 Sep 2008 11:01:14 -0700
> From: "Matt Lebo" <lebo at="" usc.edu="">
> Subject: [BioC] Missing Values in contrasts.fit
> To: bioconductor at stat.math.ethz.ch
>
> Hi,
> I am working on a time course series looking at expression
differences
> between males and females. We expect some genes to be sex-
differentially
> expressed at only some time points, and to show no expression at
other time
> points (i.e have missing values). When I use contrasts.fit to do an
> ANOVA-like analysis to find genes with sex-differential expression,
any gene
> like this will not be analyzed. However, using the aov function in
R, these
> genes are analyzed. Is there any way to include these genes so they
are
> analyzed when using contrasts.fit?
>
> Thanks,
> --
> Matthew Lebo
> Molecular and Computational Biology
> University of Southern California