normalizeQuantile.matrix {aroma.light}R Documentation

Weighted sample quantile normalization

Description

Normalizes channels so they all have the same average sample distributions.

The average sample distribution is calculated either robustly or not by utilizing either weightedMedian() or weighted.mean(). A weighted method is used if any of the weights are different from one.

Usage

## S3 method for class 'matrix':
normalizeQuantile(X, ties=FALSE, robust=FALSE, weights=NULL, typeOfWeights=c("channel", "signal"), ...)

Arguments

X a numerical NxK matrix with the K columns representing the channels and the N rows representing the data points.
robust If TRUE, the (weighted) median function is used for calculating the average sample distribution, otherwise the (weighted) mean function is used.
ties Should ties be specially treated or not?
weights If NULL, non-weighted normalization is done. If channel weights, this should be a vector of length K specifying the weights for each channel. If signal weights, it should be an NxK matrix specifying the weights for each signal.
typeOfWeights A character string specifying the type of weights given in argument weights.
... Not used.

Value

Returns an NxK matrix.

Missing values

Missing values are excluded when estimating the "common" (the baseline) distribution. Values that are NA before remain NA. No new NAs are introduced.

Weights

Currently only channel weights are support due to the way quantile normalization is done. If signal weights are given, channel weights are calculated from these by taking the mean of the signal weights in each channel.

Author(s)

Adopted from Gordon Smyth (http://www.statsci.org/) in 2002 & 2006. Original code by Ben Bolstad at Statistics Department, University of California. Support for calculating the average sample distribution using (weighted) mean or median was added by Henrik Bengtsson (http://www.braju.com/R/).

See Also

median, weightedMedian(), mean() and weighted.mean.

Examples

# Simulate three samples with on average 20% missing values
N <- 10000
X <- cbind(rnorm(N, mean=3, sd=1),
           rnorm(N, mean=4, sd=2),
           rgamma(N, shape=2, rate=1))
X[sample(3*N, size=0.20*3*N)] <- NA

# Normalize quantiles
Xn <- normalizeQuantile(X)

# Plot the data
layout(matrix(1:2, ncol=1))
xlim <- range(X, Xn, na.rm=TRUE);
plotDensity(X, lwd=2, xlim=xlim, main="The three original distributions")
plotDensity(Xn, lwd=2, xlim=xlim, main="The three normalized distributions")

[Package aroma.light version 1.6.0 Index]