[BioC] Dup correlation in single-channel analysis?- technical rep

[Gordon K Smyth]

Dear Heather,

You can't combine blocking and single-channel analysis in limma, at least not yet.

Also, the single channel analysis is done by lmscFit() not lmFit(), hence the error message.

Best wishes
Gordon

> Date: Mon, 02 Apr 2007 10:21:38 -0700
> From: Heather de Glanville <deglanvi at pdx.edu>
> Subject: [BioC] Dup correlation in single-channel analysis?- technical
> 	rep
> To: bioconductor at stat.math.ethz.ch
>
> Dear Bioconductor,
>
> I am working with a loop design with both biological and technical
> replication, and am having difficulty incorporating both types of
> replication appropriately. I am looking at the both strain and the
> treatment level effects, will make most all possible contrasts, and
> would like to test this over the biorep error.
>
> My experiment follows, with 6 different strain/treatments
> (CD,CW,HD,HW,VD,VW), 3 bio reps per strain/treat (1,2,3) and 2 techreps
> per biorep.
>
[targets frame deleted]
>
> It seems like I could use single-channel analysis, and use duplicate
> correlation to indicate which file and channels are the technical reps.
> I am getting an error when attempting to do this
>
> targets2 <- targetsA2C(targets)
> u <- unique(targets2$Target)
> f <- factor(targets2$Target, levels=u)
> design <- model.matrix(~0+f)
> colnames(design) <- u
> biolrep
> <-c(11,18,7,11,10,8,8,15,9,1,12,4,13,17,16,14,14,3,17,6,18,10,15,9,1,5,2,7,4,2,5,16,3,13,6,12)
> corfit <-duplicateCorrelation(MAAq, ndups=1, block=biolrep)
> fit <- lmFit(MAAq, block=biolrep, cor=corfit$consensus)
>
> Error in gls.series(y, design = design, ndups = ndups, spacing = spacing, :
>
> Length of block does not match number of arrays
>
> Thank you for looking over this case. I?m struggling with the R program
> in general, so perhaps that accounts for some of my general confusion.
>
> Sincerely,
>
> Heather de Glanville