Row-by-Row Linear-Model Residuals for Gene Expression (or similar) Data Structures

Share:

Description

This produces residuals of an identical linear model applied to each row of a gene expression matrix (or similar dataset). Computation speed is achieved via straightforward matrix algebra. Most commonly-used residual types are available.

Usage

1
getResidPerGene(lmobj, type = "extStudent")

Arguments

lmobj

An object produced by function lmPerGene.

type

A string indicating the type of residual requeseted (defaults to externally-Studentized).

Details

Types of residuals now available:

"response"

Response residuals, observed minus fitted

"normalized"

Response residuals divided by the estimated residual S.E.

"intStudent"

Internally Studentized residuals, often referred to as "Standardized"

default

Externally Studentized residuals, which can be used directly for outlier identification

Value

Returns a instance of ExpressionSet where the expression matrix contains the residuals. The phenoData are inherited from lmobj$eS.

Author(s)

Robert Gentleman, Assaf Oron

See Also

lmPerGene, resplot,dfbetasPerGene,influence.measures

Examples

1
2
3
4
5
6
7
8
data(sample.ExpressionSet)
lm1 = lmPerGene(sample.ExpressionSet,~sex)
r1 = getResidPerGene(lm1)
### now a boxplot of all residuals by sample
resplot(resmat=exprs(r1),fac=sample.ExpressionSet$sex)
### This plot is not very informative because of some gross outliers;
### try this instead
resplot(resmat=exprs(r1),fac=sample.ExpressionSet$sex,lims=c(-5,5))