[BioC] limma: duplicateCorrelation error

[Gordon K Smyth]

> Date: Fri, 18 Feb 2005 13:20:00 +0100
> From: Georg Otto <georg.otto at tuebingen.mpg.de>
> Subject: [BioC] limma: duplicateCorrelation error
> To: bioconductor at stat.math.ethz.ch
>
> Dear Bioconductors,
>
> I am working with an array containing 17496 probes spotted in
> duplicate. To deal with duplicate spots I used duplicateCorrelation
> like this:
>
>   cor<-duplicateCorrelation(MA,ndups=2, spacing=1458)
>
> This results in a warning message:
>
> Warning message:
> Too much damping - convergence tolerance not achievable in:
> glmgam.fit(dx, dy, s
> tart = start, tol = tol, maxit = maxit, trace = trace)
>
> Despite this, I get a result out of it:
>
>   cor$consensus.correlation
> [1] 0.60528
>
> which i use like this:
>
>   fit<-lmFit(MA, ndups=2, correlation=0.60528)
>   fit<-ebayes(fit)

There are warnings and there are errors.  This is not an error as you've said in the subject line,
it is a warning.

This question has been asked quite a few times on this list.  As I've said each time, it is not a
problem.  duplicateCorrelation() take the consensus of a large number of correlations, and it
makes virtually no difference that one correlation out of thousands could not be computed.

Gordon

> writing the M values from fit to a table, I notice that some probes are
> missing from the output. I looked at differences to a fit that was done
> without duplicateCorrelation trying to rule out that there is a problem
> with the spacing of probes. I looked at three consecutive probes that
> are duplicated with the same spacing, but after duplicateCorrelation
> the middle one is missing while the two other ones are there in
> duplicates. This leaves me clueless...
>
> Any idea?
>
> Best,
>
> Georg
>
> --
> Georg Wilhelm Otto
> Max-Planck-Institute for Developmental Biology
> Spemannstrasse 35/III
> D-72076 Tuebingen Germany
> phone: +49-7071-601 301
> http://www.eb.tuebingen.mpg.de