[BioC] NormalizeBetweenArrays

[Gordon Smyth]

At 10:43 PM 11/03/2004, Jason Skelton wrote:
>Gordon Smyth wrote:
>>This isn't built-in to limma but it is easy to do using the subset and 
>>cbind operations defined for MAList objects. Suppose that 'design' is the 
>>matrix you give above and 'MA' is your within-array normalized data object.
>
>Thankyou very much for this Gordon this is exactly what I wanted......
>On a different note
>
>since I changed to newer version of limma (1.5.1)
>I receive the following message when I run dupcor.series or 
>duplicateCorrelation
>
>duplicateCorrelation(object,design=rep(1,ncol(M)),ndups=2,spacing=1,trim=0.15,weights=NULL)
>dupcor.series(M,design=rep(1,ncol(M)),ndups=2,spacing=1,initial=0.8,trim=0.15,weights=NULL)
>
>
>>NBAScaleMULTIcor <- dupcor.series(NBAScaleMULTI$M, SchistodesignMATRIX, 
>>ndups=2, spacing=544)
>
>There were 12 warnings (use warnings() to see them)
> > warnings()
>Warning messages:
>1: Max iterations exceeded in: glmgam.fit(dx, dy, start = c(mean(dy), 0))
>2: Max iterations exceeded in: glmgam.fit(dx, dy, start = c(mean(dy), 0))
>3: Max iterations exceeded in: glmgam.fit(dx, dy, start = c(mean(dy), 0))
>4: Max iterations exceeded in: glmgam.fit(dx, dy, start = c(mean(dy), 0))
>5: Max iterations exceeded in: glmgam.fit(dx, dy, start = c(mean(dy), 0))
>6: Max iterations exceeded in: glmgam.fit(dx, dy, start = c(mean(dy), 0))
>7: Max iterations exceeded in: glmgam.fit(dx, dy, start = c(mean(dy), 0))
>8: Max iterations exceeded in: glmgam.fit(dx, dy, start = c(mean(dy), 0))
>9: Max iterations exceeded in: glmgam.fit(dx, dy, start = c(mean(dy), 0))
>10: Max iterations exceeded in: glmgam.fit(dx, dy, start = c(mean(dy), 0))
>11: Max iterations exceeded in: glmgam.fit(dx, dy, start = c(mean(dy), 0))
>12: Max iterations exceeded in: glmgam.fit(dx, dy, start = c(mean(dy), 0))
>
>Any ideas ?

Just ignore this - it isn't a problem. I have replaced the call to gls() 
from the nlme library with my own program randomizedBlock from the statmod 
library, because it is quite a bit faster on typical microarray problems. 
It is the new program which is generating the warnings.

Gordon

>cheers
>Jason