[BioC] summarizing probe intensites before or after normalization- 1. how to do with RMA 2. Opinions?

[Ben Bolstad]

On Mon, 2006-09-11 at 16:44 +0200, k. brand wrote:
> Ben,
> 
> Thankyou for your fast response.
> 
> I tried your suggested script as:
> 
> library(affyPLM)
> dat <- ReadAffy()
> datrma <- rma(dat, normalize=FALSE)
> datrma.postqnorm <- normalize(datrma)
> boxplot(datrma)
> 

try 

boxplot(datrma.postqnorm)


Here is my test code that shows it works

library(affyPLM)
data(Dilution)
rma.nonorm <- rma(Dilution,normalize=FALSE)
rma.postnorm <- normalize(rma.nonorm)
boxplot(rma.nonorm)
boxplot(rma.postnorm)


> Im convinced RMA is a superior approach to MAS5. The variation of the 
> spreads however, compared to normalising in the last stage is 
> surprising. From "RMA no norm.jpeg" you can see my data is quite 
> 'divergent' which is exactly why i want to normalize as best i can. But 
> perhaps RMA is not 'strong' enough to push around such divergence?

Based on the plot you sent me:

1) is there a biological reason for the pattern I see (ie first two low,
next two high, next two low ....) or is this some sort of technical
artifact.

2) Have you carried out any quality assessment?