[BioC] median normalization in limma

[Gordon K Smyth]

Dear S Fazeli,

Median normalization does not affect the spread of the M-values within 
each array, but it does affect the spread of the M-values between arrays. 
The whole purpose of normalization is to make different arrays more alike.

You would be better off displaying boxplots by array, as illustrated in 
the limma User's Guide for the Swirl data set.

Best wishes
Gordon

> Date: Sat, 18 Jun 2011 08:54:33 +0430
> From: samane fazeli <samanefazeli at gmail.com>
> To: <bioconductor at stat.math.ethz.ch>
> Subject: [BioC] median normalization in limma
> Message-ID: <BANLkTinJ3WyRAYK4sLrcTNyX6RSd8+FgFA at mail.gmail.com>
> Content-Type: text/plain
>
> Dear all
> As we know a median normalization shifts the center of the log-ratio
> distribution to zero but does not affect the spread.
> when i plot a density plot after median normalization, impact of median
> normalization on spread of data is observed.
> you can observe the density plot in R as follows
>
> library(limma)
> targets=readTargets("SwirlSample.txt")
> RG=read.maimages(targets$FileName, source="spot")
>
> MA=normalizeWithinArrays(RG,method="none")
> M1=MA$M
> plot(density(M1,bw=0.17),main="density plots",xlab="log ratios",
> ylab="density",col=1,lty=1,lwd=3)
> MA=normalizeWithinArrays(RG,method="median")
> M2=MA$M
> lines(density(M2,bw=0.17),col=8,lty=1,lwd=3)
>
> I will be appreciate if you could me any hint about this issue
> Thank you in advance
>
> Regards
>
>
> -- 
> *S. Fazeli
> MSc Student of Chem. Eng.
> ferdowsi University
> mashhad
> Iran*

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}