[BioC] question on decideTests function in limma

[Gordon K Smyth]

> Date: Mon, 08 May 2006 17:17:50 -0400
> From: CHUNFA C JIE <cjie1 at jhmi.edu>
> Subject: [BioC] question on decideTests function in limma
> To: bioconductor at stat.math.ethz.ch
> Message-ID: <27808c9c2329e.445f7d3e at jhmimail.jhmi.edu>
> Content-Type: text/plain; charset=us-ascii
>
> Hi,
>
> With limma, the moderated-t and its p-value provide a ranking
> guide because the various sorts of underlying assumptions are never exactly true
> for microarray data.
>
> The decideTests() fucntion makes multiple testing adjstment and declares significance.
>
> My question:
>       Dr. Smyth says the moderated-t and its p-value "are intended to be taken as a guide rather
> than
> as a strict measure of the probability of differential expression".  Then how does decideTests()
> fuction
> declare significance (yielding -1, 0, 1)?  Are the underlying assumptions assumed when using
> decideTests()?
>
>
>
>
> Chunfa (Charles) Jie,  Ph.D./ M.S.
> JHMI Microarray Core
> School of Medicine
> Johns Hopkins University
> BRB 355, 733 W Broadway
> Baltimore, MD 21205
> Phone: (410) 502-6659
> Fax:      (410)502-7543
> Email: cjie1 at jhmi.edu

Dear Charles,

The phrase you quote refered to the B-statistics, not to the moderated-t and its p-value.

Best wishes
Gordon