permute: Permutation test

In ClusterSignificance: The ClusterSignificance package provides tools to assess if class clusters in dimensionality reduced data representations have a separation different from permuted data

Description

Test how the classification performs compared to random (eg. permuted) data.

Usage

## S4 method for signature 'PermutationResults'
getData(x, n = NULL)

## S4 method for signature 'PermutationResults'
c(x, ..., recursive = FALSE)

pvalue(x, ...)

## S4 method for signature 'PermutationResults'
pvalue(x, ...)

conf.int(x, ...)

## S4 method for signature 'PermutationResults'
conf.int(x, conf.level = 0.99, ...)

## S4 method for signature 'PermutationResults'
initialize(.Object, ..., scores.real, scores.vec)

permute(mat, ...)

## S4 method for signature 'matrix'
permute(mat, classes, projmethod = "pcp", iter = 100,
  user.permutations = NULL, seed = 3, df = NULL, verbose = TRUE, ...)

## S4 method for signature 'PermutationResults,missing'
plot(x, y, comparison = "all", ...)

## S4 method for signature 'PermutationResults'
show(object)

Arguments

x	matrix for the function permute, otherwise it is a PermutationResults object
n	data to extract from ClassifiedPoints (NULL gives all)
...	arguments to pass on
recursive	dont use (belongs to default generic of combine 'c()')
conf.level	confidence level for the returned confidence interval
.Object	internal object
scores.real	the real score
scores.vec	all permuted scores
mat	matrix with samples on rows, PCs in columns. Ordered PCs, with PC1 to the left.
classes	vector in same order as rows in matrix
projmethod	'pcp' or 'mlp'
iter	integer number of iterations to be performed.
user.permutations	user defined permutation matrix
seed	random seed to be used by the internal permutation
df	degrees of freedom, passed to smooth.spline
verbose	makes function more talkative
y	default plot param, which should be set to NULL
comparison	Specify a comparison i.e. ("grp1 vs grp2") and plot only that comparison.
object	ClassifiedPoints Object

Details

This is a test suit and will return a summarized object. The default of the parameter 'iter' is set quite low, and in principle the more iterations the better, or until the pvalue converges to a specifc value. If no pre-permuted data has been supplied by the user, then the internal permutation method will perform a sampling without replacement within each dimension.

Value

The permute function returns an object of class PermutationResults

Author(s)

Jesper R. Gadin and Jason T. Serviss

Examples

#use pcp method
data(pcpMatrix)
classes <- rownames(pcpMatrix)

#run function
iterations <- 10
pe <- permute(
    mat=pcpMatrix,
    classes=classes,
    iter=iterations,
    projmethod="pcp"
)

#use mlp method
data(mlpMatrix)
classes <- rownames(mlpMatrix)
pe <- permute(
    mat=mlpMatrix,
    classes=classes,
    iter=iterations,
    projmethod="mlp"
)


#getData accessor
getData(pe)

#getData accessor specific
getData(pe, "scores.vec")

#get pvalue
pvalue(pe)

#plot result
plot(pe)

#combine three (parallell) jobs on the same matrix 
pe2 <- c(pe, pe, pe)

ClusterSignificance documentation built on Nov. 8, 2020, 5:28 p.m.