[BioC] Choosing results from LIMMA analysis

[Gordon K Smyth]

Dear Sharon,

You are asking to be told that one method is always best, but if there was such a method then the
alternatives wouldn't be provided in the software.  Different methods provide better power against
different types of alternatives.  You might start by searching the mailing list archives for
comments on the different methods.

Best wishes
Gordon


Original message:
[BioC] Choosing results from LIMMA analysis
Sharon sharonanandhi at gmail.com
Wed Aug 2 14:32:08 CEST 2006

Hi All,

I have used limma to analyze the time course microarray data (ctrl, 3Hr,
12Hr and 24Hr; each having 3 arrays). Using eBayes(), I have calculated
F.p.value and it's corresponding FDR using p.adjust. I also used
'decideTests(fit1, method="nestedF", adjust.method="BH")'.
Now I can choose probesets based on FDR calculated by p.adjust or by using
results given by 'nestedF'. I can clearly see that the number of significant
probesets given by 'nestedF' is larger than by 'FDR calculated by p.adjust'
at the same cutoff value 0.05. Which one is the preferred method for
choosing significant probesets?

Your suggestions are greatly appreciated.

Thanks,
Sharon