[BioC] Limma: design for a multiple level factor taking into account centre and plate

[Daniel Brewer]

Hello,

We have a microarray dataset where there seems to be significant
variation caused by the centre (2 centres) the sample came from and the
plate (2 plates), so I would like to basically remove these effects by
including them in the linear model for any other comparison I do.  I
have a factor sampleType that has levels 1 to 4 and I am interested in
the comparisons 1 vs 2, 1 vs 3 and 1 vs 4.  Is this a suitable way to
set up the design matrix:

design <- model.matrix(~centre + plate + sampleType,data=data)

which produces something like:
          (Intercept) Plate2 CentreRMH sampleType2 sampleType3 samType4
Cb016_001           1      0         0           1           0        0
Cb016_003           1      0         1           0           0        1
Cb016_004           1      0         1           0           0        1

I then look at the top table for coefficient 4 to get the result of
sample type 2 vs sample type 1

Is this a reasonable way to do it?  The reason I am unsure is that to
test it I reduced the dataset to those samples just containing
sampleType 1 and 2 and got different results, but this could be
explained by different estimates for centre and plate and a different
estimate produced by eBayes.

Many Thanks

-- 
**************************************************************
Daniel Brewer, Ph.D.

Institute of Cancer Research
Molecular Carcinogenesis
Email: daniel.brewer at icr.ac.uk
**************************************************************

The Institute of Cancer Research: Royal Cancer Hospital, a charitable Company Limited by Guarantee, Registered in England under Company No. 534147 with its Registered Office at 123 Old Brompton Road, London SW7 3RP.

This e-mail message is confidential and for use by the a...{{dropped:2}}