normalizeQuantileRank: Normalizes the empirical distribution of one of more samples...

In HenrikBengtsson/aroma.light-BioC_release: Light-Weight Methods for Normalization and Visualization of Microarray Data using Only Basic R Data Types

Description

Normalizes the empirical distribution of one of more samples to a target distribution.

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 'numeric'
normalizeQuantileRank(x, xTarget, ties=FALSE, ...)
## S3 method for class 'list'
normalizeQuantileRank(X, xTarget=NULL, ...)
## Default S3 method:
normalizeQuantile(x, ...)

Arguments

x, X	a numeric vector of length N or a list of length N with numeric vectors. If a list, then the vectors may be of different lengths.
xTarget	The target empirical distribution as a sorted numeric vector of length M. If NULL and X is a list, then the target distribution is calculated as the average empirical distribution of the samples.
ties	Should ties in x be treated with care or not? For more details, see "limma:normalizeQuantiles".
...	Not used.

Value

Returns an object of the same shape as the input argument.

Missing values

Missing values are excluded when estimating the "common" (the baseline). Values that are NA remain NA after normalization. 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.

See Also

To calculate a target distribution from a set of samples, see averageQuantile(). For an alternative empirical density normalization methods, see normalizeQuantileSpline().

Examples

# Simulate ten samples of different lengths
N <- 10000
X <- list()
for (kk in 1:8) {
  rfcn <- list(rnorm, rgamma)[[sample(2, size=1)]]
  size <- runif(1, min=0.3, max=1)
  a <- rgamma(1, shape=20, rate=10)
  b <- rgamma(1, shape=10, rate=10)
  values <- rfcn(size*N, a, b)

  # "Censor" values
  values[values < 0 | values > 8] <- NA

  X[[kk]] <- values
}

# Add 20% missing values
X <- lapply(X, FUN=function(x) {
  x[sample(length(x), size=0.20*length(x))] <- NA;
  x
})

# Normalize quantiles
Xn <- normalizeQuantile(X)

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

HenrikBengtsson/aroma.light-BioC_release documentation built on May 7, 2019, 1:55 a.m.