RankingSoftthresholdT {GeneSelector}R Documentation

Ranking via the 'soft-threshold' t-statistic

Description

The 'soft-threshold' statistic is constructed using a linear regression model with the L1 penalty (also referred to as LASSO penalty). In special cases (like here) the LASSO estimator can be calculated analytically and is then called 'soft threshold' estimator (Wu,2005).
For S4 method information, see RankingSoftthresholdT-methods.

Usage

RankingSoftthresholdT(x, y, type = c("unpaired", "paired", "onesample"),
                      lambda = c("lowess", "cor", "user"), userlambda = NULL, 
                      gene.names = NULL, ...)

Arguments

x A matrix of gene expression values with rows corresponding to genes and columns corresponding to observations or alternatively an object of class ExpressionSet.
If type = paired, the first half of the columns corresponds to the first measurements and the second half to the second ones. For instance, if there are 10 observations, each measured twice, stored in an expression matrix expr, then expr[,1] is paired with expr[,11], expr[,2] with expr[,12], and so on.
y If x is a matrix, then y may be a numeric vector or a factor with at most two levels.
If x is an ExpressionSet, then y is a character specifying the phenotype variable in the output from pData.
If type = paired, take care that the coding is analogously to the requirement concerning x
type
"unpaired":
two-sample test.
"paired":
paired test. Take care that the coding of y is correct (s. above)
"onesample":
y has only one level. Test whether the true mean is different from zero.

lambda s. details
userlambda A user-specified value for lambda, s. details.
gene.names An optional vector of gene names.
... Currently unused argument.

Details

There are currently three ways of specifying the shrinkage intensity lambda. Both "lowess" and "cor" are relatively slow, especially if rankings are repeated (GetRepeatRanking). Therefore, a 'reasonable' value can be set by the user.

Value

An object of class GeneRanking.

Note

The code is a modified version of that found in the st package of Opgen-Rhein and Strimmer (2007).

Author(s)

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

References

Wu, B. (2005). Differential gene expression using penalized linear regression models: The improved SAM statistic. Bioinformatics, 21, 1565-1571

See Also

GetRepeatRanking, RankingTstat, RankingFC, RankingWelchT, RankingWilcoxon, RankingBaldiLong, RankingFoxDimmic, RankingLimma, RankingEbam, RankingWilcEbam, RankingSam, RankingBstat, RankingShrinkageT, RankingPermutation, RankingGap

Examples

### Load toy gene expression data
data(toydata)
### class labels
yy <- toydata[1,]
### gene expression
xx <- toydata[-1,]
### run RankingSoftthresholdT
softt <- RankingSoftthresholdT(xx, yy, type="unpaired")

[Package GeneSelector version 1.4.0 Index]