[BioC] ANCOVA in limma

[Gordon Smyth]

At 08:00 PM 20/10/2005, bioconductor-request at stat.math.ethz.ch wrote:
>Date: Wed, 19 Oct 2005 14:26:42 -0400
>From: Naomi Altman <naomi at stat.psu.edu>
>Subject: [BioC] ANCOVA in limma
>To: bioconductor at stat.math.ethz.ch
>
>We would like to adjust our ANOVA-type analysis by using the
>covariate "age".  Because limma is based on contrast, rather than
>"type III SS", it seems to me that the only way to proceed with an
>ANCOVA-type analysis is to orthogonalize the design matrix against
>the covariate.  Is there a more straight-forward method?

I think that the limma approach is essentially equivalent to SAS Type III 
Anova. I think you should get the same results from orthogonalizing your 
design matrix as you'd get from the standard limma analysis, in which you 
include age in the design matrix but not in your contrasts. If you find 
this isn't the case, let me know.

Gordon

>Naomi S. Altman                                814-865-3791 (voice)
>Associate Professor
>Dept. of Statistics                              814-863-7114 (fax)
>Penn State University                         814-865-1348 (Statistics)
>University Park, PA 16802-2111