[BioC] modelMatrix question

[Gordon K Smyth]

You're not using the 'parameters' argument to modelMatrix() correctly.  The main point though is
that there is no need to take special action to allow for different numbers of replicates.  This
is accommodated automatically in the course of an ordinary analysis.

Gordon


> Date: Wed, 18 May 2005 13:49:33 +0100
> From: Jason Skelton <jps at sanger.ac.uk>
> Subject: [BioC] modelMatrix question
> To: bioconductor at stat.math.ethz.ch
>
> Hi
> I have a question regarding the modelMatrix function in limma.
> Which has been covered before however I haven't found an appropriate answer in the
> documentation or the list archives.
>
> Can one use modelMatrix to adjust for unequal numbers of technical
> replicates? To take a very simplified example for:
>
> targetsx
> 	Cy3	Cy5
> 1	WT_sample1	Mutant_sample1
> 2	Mutant_sample1	WT_sample1
> 3	WT_sample2	Mutant_sample2
> 4	Mutant_sample3	WT_sample3
>
> The samples being different biological replicates of single mutant and
> wild-type cell lines. So we would just want to average them together, but
> without biassing towards sample1 relatively to samples 2 and 3.
>
> I thought I might be able to enter parameters like
>
> parameters
> 		Mutant-WT
> WT_sample1	-1/3
> WT_sample2	-1/3
> WT_sample3	-1/3
> Mutant_sample1	1/3
> Mutant_sample2	1/3
> Mutant_sample3	1/3
>
> to give the different replicates equal weight.
> But this just gives me
>
>
>>> modelMatrix(targetsx, parameters)
>>
>>
> Found unique target names:
>  Mutant_sample1 Mutant_sample2 Mutant_sample3 WT_sample1 WT_sample2
> WT_sample3
>   [,1]
> 1    1
> 2   -1
> 3    1
> 4   -1
>
> which doesn't appear to be what I want
> any help greatfully recieved
>
> Jason
>
> --
> --------------------------------
> Jason Skelton
> Pathogen Microarrays
> Wellcome Trust Sanger Institute
> Hinxton
> Cambridge
> CB10 1SA
>
> Tel +44(0)1223 834244 Ext 7123
> Fax +44(0)1223 494919
>
>
>
> ------------------------------