Dear list,
I obtain different values for chip effects
using fitPLM or rmaPLM:
>Pset<-fitPLM(Data,model=PM~-1+probes+samples,output.param=list(weight
s=TRUE))
>Pset.rma <-rmaPLM(Data,output.param=list(weights=TRUE))
I did not expect that as I thought that, by default, both procedure
use the
same bkg+normalization+summarization
Any help will be welcome
Regards,
Linux AMD Opteron 64bit
R Version 2.0.1
affyPLM 1.2.5
affy1.5.8
Ariel./
--
Ariel Chernomoretz, Ph.D.
Centre de recherche du CHUL
2705 Blv Laurier, bloc T-367
Sainte-Foy, Qc
G1V 4G2
(418)-525-4444 ext 46339
The bkg and normalization routines in your calls are identical, the
difference lies in the summarization algorithm.
rmaPLM() uses the median polish so that the chip effects returned are
identical to the values you get out of the rma() function. However,
rmaPLM() returns a PLMset object, which means it is possible to get
the
resulting residuals and probe-effect coefficients estimates. Note that
the weights returned from rmaPLM() are synthetic (ie not part of the
modeling procedure) but may satisfactorily be used for visualization.
fitPLM() uses robust regression for the model fitting procedure. The
weights returned are the weights used in the final stage of the
iterative reweighted least squares fitting algorithm.
Ben
On Wed, 2005-05-04 at 16:02 -0400, Ariel Chernomoretz wrote:
> Dear list,
>
> I obtain different values for chip effects
> using fitPLM or rmaPLM:
>
> >Pset<-fitPLM(Data,model=PM~-1+probes+samples,output.param=list(weig
hts=TRUE))
> >Pset.rma <-rmaPLM(Data,output.param=list(weights=TRUE))
>
> I did not expect that as I thought that, by default, both procedure
use the
> same bkg+normalization+summarization
>
> Any help will be welcome
> Regards,
>
> Linux AMD Opteron 64bit
> R Version 2.0.1
> affyPLM 1.2.5
> affy1.5.8
>
>
> Ariel./
>
>