[BioC] duplicateCorrelation with few arrays

[Gordon K Smyth]

Dear Sergio,

The duplicateCorrelation document tells you that you need 2 more arrays than coefficients defined
by the design matrix, and I think that you have already read that.  You are, in effect, asking
whether the documentation is correct.  It is.  Getting a correlation of 1 is exactly the sort of
thing which will happen if you don't have the minimum number of arrays.

Best wishes
Gordon

> Date: Wed, 21 Feb 2007 10:31:16 -0800
> From: Sergio Barberan <barberan at biology.ucsc.edu>
> Subject: [BioC] duplicateCorrelation with few arrays
> To: bioconductor at stat.math.ethz.ch
> Message-ID: <1172082676.10408.234.camel at elegans.ucsc.edu>
> Content-Type: text/plain
>
> Hi,
>
> It is possible that my only problem is that I'm trying to use
> duplicateCorrelation with few arrays but I just wanted to be sure.
>
> My slides include three replicas of each spot 120 spots apart;
> to make sure I'm giving duplicateCorrelation the right input I do:
>
> unwrapdups(MA$genes$Name, ndups=3, spacing=120)
>
> and I do obtain a matrix with identical columns with the replica spots
> on it.
>
> My problem is that when I try to calculate:
>
> cor <- duplicateCorrelation(MA,design,ndups=3,spacing=120)
>
> I get several warnings: 1: NaNs produced in: sqrt(dfitted.values)
>
> and I get a correlation of 1.
>
> Right now I do not have two more arrays than the column rank of design,
> but I was still trying to get some results; is this the problem or is
> there something I'm doing wrong.
>
> Thanks in advance for your help,
> Sergio