[BioC] Comparing limma results of two different groupings

Kasper Daniel Hansen khansen at stat.Berkeley.EDU
Mon Aug 6 00:34:26 CEST 2007


On Aug 3, 2007, at 8:01 AM, Daniel Brewer wrote:

> I have been using limma to do differential analysis between two
> different groupings, both splitting the samples into two groups.  The
> results for one set give approx. 100 significant genes (adj.
> p-value<0.05) whereas the other gives approx. 1000 significant genes.
> Is this suggestive that the second grouping is more "accurate" than  
> the
> other?  What is the best way to test this in a statistical manner.

Why do you think the results are more accurate / better, just because  
you have more DE genes. The best result is the one that is closest to  
the true list of DE genes when taking the experimental . Also, if the  
two groupings are different, it might very well be that both lists  
are true at the same time. It is hard to comment on this without  
knowing more about what you are doing. And I would be hard pressed to  
come up with a situation where more DE is necessarily better (in the  
"more true" sense as opposed to the "I would really like a big list"  
sense).

Kasper



More information about the Bioconductor mailing list