Thanks Mark,
Using the core cdfs for both human and mouse I still get a few
probesets with NaN values but indeed they have 1 probe per probeset
only. I was wondering also from what probe values the MAD statistic is
computed that is used to scale the residuals and compute the FIRMA
scores. The FIRMA paper does not say this clearly. Is it from the exon
probes or is it from the transcript probes.
I have seen the message and I can see
there that the scaling factor (variable u.mad) is a constant for the
entire dataset (whole samples and whole transcripts and whole exons)
therefore I am not sure I understand how this scaling makes the scores
comparable between exons and transcripts.


On Aug 5, 9:34 pm, Mark Robinson <> wrote:
> Hi Adi.
> First, you seem to be doing the FIRMA analysis on probeset-level data (i.e. 
> roughly 4 probes per exon), using the fullR1,A20080718,MR.  FIRMA is really a 
> method that uses all of the probe-level data for an entire *gene* and then 
> summarizes residuals to score differential splicing at the probeset level.  
> You'll therefore want to use a CDF file that is oriented this way (perhaps 
> MoEx-1_0-st-v1,coreR1,A20080718,MR.cdf or 
> MoEx-1_0-st-v1,U-Ensembl50,G-Affy,EP.cdf?)
> I think the NaNs you get are just from the fact that some exons only have a 
> single probe (and, as you say, group=1 below!).
> Cheers,
> Mark

When reporting problems on aroma.affymetrix, make sure 1) to run the latest 
version of the package, 2) to report the output of sessionInfo() and 
traceback(), and 3) to post a complete code example.

You received this message because you are subscribed to the Google Groups 
"aroma.affymetrix" group with website
To post to this group, send email to
To unsubscribe and other options, go to

Reply via email to