[BioC] limma: print-tip loess and empty spots

[Gordon Smyth]

Dear Adrian,

Raw M-values are always defined to be log(Red/Green), i.e., 635/532. 
limma always computes them like that, because that's the definition. 
If you think you need to change this, you're almost certainly not 
thinking about your data in the best possible way.

Correct allowance for which treatments have been labelled red and 
green is done at a later stage, not at the normalization stage. 
limmaGUI provides plenty of facilities to do this, but if you export 
the data out of the limma to treat it using other tools, then 
obviously this becomes your responsibility.

BTW, limmaGUI provides features for checking data quality, such as 
MA-plots and image plots.

Best wishes
Gordon

>Date: Wed, 30 May 2007 15:27:33 -0700
>From: "Adrian Steward" <adrian.steward0405 at gmail.com>
>Subject: Re: [BioC] limma: print-tip loess and empty spots
>To: keith at wehi.edu.au, bioconductor at stat.math.ethz.ch
>
>Hi Keith et al.,
>
>Thanks for replying, I've been trying to learn limma as I go.  I haven't
>gotten into the linear modeling portion, but am rather using limma to
>pre-process (e.g. normalize) my data to evaluate its quality before I get
>serious with limma GUI for contrasts, heat maps, and all the nice stuff it
>can do.  Thus far, I have found no way to change the default setting for
>635/532 - regardless of how the data is read in, the ratios are calculated
>this way by limma.
>
>Is there a bit of code I can enter into the command line to change this?
>All suggestions are welcome.
>
>Thanks!
>
>AM