upperBoundNull: function to estimate upper limit of null distribution

In Ringo: R Investigation of ChIP-chip Oligoarrays

Description

The data is assumed to arise from a mixture of two distributions, a symmetric null distribution with its mode close to zero, and an alternative distribution that is stochastically larger than the null. This function tries to pinpoint the minimum of data values that are more likely to arise from the alternative distribution, i.e. an upper bound for values following the null distribution.

Usage

upperBoundNull(x, modeMethod = "shorth", limits = c(-1, 1), prob = 0.99, ...)

Arguments

x	numeric vector of data values
modeMethod	character string; which method to use for estimating the mode of the null distribution; see details
limits	numeric of length 2; data values within this range are used for estimating the mode of the null distribution
prob	quantile of the null distribution that will be used as an upper bound
...	additional arguments that are passed on to the function for mode estimation

Details

For estimating the mode of the null distribution, current options are

"shorth"

the function shorth

"half.range.mode"

the function half.range.mode

"null"

does not estimate the mode from the data, but sets it to 0

Value

a single numeric value which is the estimated upper bound for the null distribution.

Note

This way of estimating the null distribution is mentioned in the PhD thesis of Richard Bourgon.

Author(s)

Joern Toedling, based on suggestions by Richard Bourgon

See Also

shorth, half.range.mode

Examples

  exDir <- system.file("exData",package="Ringo")
  load(file.path(exDir,"exampleProbeAnno.rda"))
  load(file.path(exDir,"exampleX.rda"))
  smoothX <- computeRunningMedians(exampleX, probeAnno=exProbeAnno,
       modColumn = "Cy5", allChr = "9", winHalfSize = 400)
  apply(exprs(smoothX), 2, upperBoundNull)

Ringo documentation built on Nov. 8, 2020, 5:34 p.m.