[BioC] NA and quantile normalization
bolstad at stat.berkeley.edu
Fri Jul 22 03:12:20 CEST 2005
If your data initially has NAs then it is quite probable that
normalize.quantiles() would introduce additional NA (or more likely turn
the entire matrix into NA) since it does not make special allowances for
NA values. I think the normalizeQuantiles() function in the limma
package handles NA's more appropriately.
On Thu, 2005-07-21 at 13:59 -0500, Ravi Murthy wrote:
> I am new to R,
> I am doing quantile normalization with a data
> consisting of a matix 384 X 124.
> While using "normailze.quantiles" I find that it introduces 'NA' into some
> of the
> cells, where we should expect a numerical value, can someone help me to
> overcome this problem ?
> The matrix consists of 62 plates in repliccate, and
> quantile normailization is done for the data from replicates (colomn 1,2)
> using normailze.quantiles function from 'affy' package resulting in data
> output as g1
> >g1 <- normalize.quantiles(exprs(MSExpr[,1:2]))
> etc., uptp g62
> This command works well for smaller series of data, but
> while using for a larger data matrix, it introduces "NA" in the place where
> it is suppose to geneerate data . It doesn't make any
> sense because elsewhere in the colomn consisting of similar values it
> computes properly.
> In addition, if I continue to perform median
> normailzation of the quantlies, the function fails to
> give any output, execept for those colomns which does
> not have any "NA". I cannot get rid of interference of NA in median
> normalization by using any of NA related functions.
> I am using R in MAC based platform for this analysis.
> Please suggest me some way to avoid 'NA' related computation problem.
> ravidmurthy at hotmail.com
> Dont just search. Find. Check out the new MSN Search!
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
More information about the Bioconductor