[BioC] Removing control probes before fitting the linear model

[Gordon K Smyth]

Dear Erika,

If you give zero weight to all values for a probe, that has the same 
effect as regards the linear model as removing the probe entirely.

Best wishes
Gordon

> Date: Mon, 20 Oct 2008 12:22:26 +0200
> From: "Erika Melissari" <erika.melissari at bioclinica.unipi.it>
> Subject: [BioC] Removing control probes before fitting the linear
> 	model
> To: <bioconductor at stat.math.ethz.ch>
> Message-ID: <006101c9329d$bff5efb0$ba517283 at maanalysis>
> Content-Type: text/plain
>
> Hello,
>
> I found out that in some examples (11.5 and 11.6) of Limma userguide it 
> is suggested the removal of the control probes from the data before 
> fitting the linear model. Usually I flag negative or positive control 
> spots as "not good" by using a wt.fun function during the loading of the 
> data (e.g. I flag witn 0 all the "not good" spots and with 1 all the 
> "good" spots) and this is because I do not use control spots in order to 
> perform normalizzation of raw data. So, I do not exclude those spots 
> explicitly before fitting the lienar model, but they receive a weight 
> that differentiate them by the rest of the data. Is it correct or I have 
> to completely eliminate the control spots?
>
> Thank you very much
>
> Erika