[BioC] normexp backgroundCorrect problem

[Gordon Smyth]

>[BioC] normexp backgroundCorrect problem
>Lance Palmer lance.palmer at stonybrook.edu
>Thu Nov 10 19:34:19 CET 2005
>
>When running
>RGb<-backgroundCorrect(RG,method="normexp")
>I get the following warning at the end.
>
>Corrected array 1
>Corrected array 2
>Corrected array 3
>Corrected array 4
>Warning message:
>Limit of numerical accuracy reached with very low intensity or very high
>background:
>setting adjusted intensity to small value in: normexp.signal(mu = out$beta +
>RG$Gb[, j], out$sigma, out$alpha,
>
>I have used normexp but not have seen this message.
>Anyone know what might be wrong?

It is hard to avoid floating point overflow or underflow in the calculations.

Try instead
 > RGb <- backgroundCorrect(RG, method="rma")
in limma 2.4.0.

This options use the RMA background method from the affy package. It's the 
same model as for normexp, just with different parameter estimates. The RMA 
method works pretty much every time.

Best wishes
Gordon