speaq (version 2.7.0)

relevant.features.p: Identify features (columns in the datamatrix) which are significantly associated with the outcome.

Description

This function produces a p-value for every column in the datamatrix, corresponding to the null hypothesis that outcome/response is independent of that feature.

Usage

relevant.features.p(
  datamatrix,
  response,
  p.adj = "BH",
  POI = 1,
  responsevector = NULL
)

Arguments

datamatrix

The data matrix with a column for each feature.

response

A vector or matrix of outcomes/responses (e.g. class labels). the length of this vector or the amount of rows in this matrix should match the amount of rows in datamatrix.

p.adj

The adjustment method for the p-values. Any of 'holm', 'hochberg', 'hommel', 'bonferroni', 'BH' (default), 'BY', 'fdr' or 'none' are accepted.

POI

Only if 'response' is a matrix! The p values of interest. This is a number indicating which column of the 'response' matrix you are interested in. POI can range from 1 (default) to the number of columns in 'response'.

responsevector

(deprecated), please use the the more general 'response' variable instead.

Value

data with the features and their (adjusted) p-values, one for every column in the datamatrix .

Examples

# NOT RUN {
nSamples <- 10
nFeatures <- 20
data.matrix <- matrix( stats::runif(n=nFeatures*nSamples, min=0,max=100), 
ncol = nFeatures, nrow = nSamples)

responseVec <- c( rep(0,nSamples/2), rep(1,nSamples/2) )
p_values <- relevant.features.p(datamatrix = data.matrix, response = 
responseVec, p.adj = 'none')
p_values_adjusted <- relevant.features.p( datamatrix = data.matrix, 
response = responseVec, p.adj = 'bonferroni')

# }