[BioC] PCA plots

[Adaikalavan Ramasamy]

Remember that points in the unlog scale is much more skewed than those
on the log scale. i.e. on the unlogged scale, most of the points are
close to zero and the ones slightly away from this cluster is much
easier to visualise. i.e. log transformation spreads the points out more
evenly.

If you points were characterised by a single gene, then something along
the following lines might be happening :

 x <- rexp( 1000 )
 par(mfrow=c(1,2))
 hist(x, main="unlogged scale")
 hist(log(x), main="log scale")

Regards, Adai



On Thu, 2005-08-18 at 11:26 +0200, Jakub Orzechowski Westholm wrote:
> Hi! I have a couple of qustions about PCA plots for microarrays. I have run a number of affymetrix arrays, and used the affy package for background correction, normalization etc (standard RMA procedure). When I then want to make PCA plot of the arrays, what makes most sense: To use the values from rma which are log scale, or to transform them back to "normal scale"? (In my case the latter gives more outliers...). Is there any established standard for doing this?
> 
> kind regards
> Jakub Orzechowski 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
>