[BioC] combining p-values for multiple genes using stouffer

[Sean Davis]

On Tue, Apr 15, 2008 at 1:01 AM, william ritchie <billthebrute at yahoo.fr> wrote:
> Dear Bioc community,
>
>
>  I have a little question about combining p-values with stouffers method.
>  One of the conditions for combining p-values is that each test be considered as independent.
>  If I was interested in a group of say, 10 genes and I wanted to know if, as a group, they were significantly over or under expressed, would it be correct to combine the p-values (calculated from a series of microarrays) for each individual gene using stouffers method to be able to conclude that this group of genes is differentially expressed ?

Hi, William.  Assuming independence among genes is problematic.  Why
not use one of the many Gene Set Enrichment approaches (PGSEA, limma
gene set test, globaltest, to name just a few).  Alternatively, you
could use permutation to get a level of significance for your
Stouffer's p-value, but I think it might be problematic to take the
Stouffer's p-value directly as the measure of significance.

Sean