AggregatePenalty {GeneSelector}R Documentation

aggregation of repeated rankings by a variance penalty

Description

The idea behind this form of aggregation is to find 'reliable' candidate genes, i.e. those ones that are highly ranked and little variable at the same time. Higher variability is stronger penalized.

Usage

AggregatePenalty(RR, lambda = NULL, k=5, theta = 50,
                     estimator = c("var", "mad", "iqr", "residuals"), ...)

Arguments

RR An object of class RepeatRanking.
lambda A positive real number, quantifying the amount of variance penalty. Default is NULL, an alternative parametrization using k and theta is used.
k Must be specified combined with theta, s.below. Not used if lambda is given.
theta A pragmatic way of finding an appropriate value for lambda is to define some threshold rank theta that is still considered relevant and some k >= 1 that expresses the impprtance of the first rank as compared to the threshold rank.
estimator The variance estimator to be used:
"var"
The usual variance estimator.
"mad"
Squared median absolute deviation.
"iqr"
Interquartile range.
"residuals"
Residuals from a multivariate regression, s. StabilityLm
... Further arguments passed to variance,RepeatRanking-method.

Value

An object of class AggregatedRanking.

Author(s)

Martin Slawski martin.slawski@campus.lmu.de
Anne-Laure Boulesteix http://www.slcmsr.net/boulesteix

See Also

AggregateSimple, AggregateBayes, AggregatePenalty, AggregatePCA


[Package GeneSelector version 1.4.0 Index]