boot.tStat: Studentized Sorensen-Dice dissimilarity statistic

In pablof1988/goSorensen: Statistical inference based on the Sorensen-Dice dissimilarity and the Gene Ontology (GO)

Studentized Sorensen-Dice dissimilarity statistic

Description

Efficient computation of the studentized statistic (^dis - dis) / ^se where 'dis' stands for the "population" value of the Sorensen-Dice dissimilarity, '^dis' for its estimated value and '^se'for the estimate of the standard error of '^dis'. Internally used in bootstrap computations.

Usage

boot.tStat(xBoot, dis)

Arguments

xBoot	either an object of class "table", "matrix" or "numeric" representing a 2x2 contingency table of joint enrichment.
dis	the "known" value of the population dissimilarity.

Details

This function is repeatedly evaluated during bootstrap iterations. Given a contingency table 'x' of mutual enrichment (the "true" dataset):

n_11	n_01
n_10	n_00,

summarizing the status of mutual presence of enrichment in two gene lists, where the subindex '11' corresponds to those GO items enriched in both lists, '01' to items enriched in the second list but not in the first one, '10' to items enriched in the first list but not enriched in the second one and '00' to those GO items non enriched in both gene lists, i.e., to the double negatives.

A typical bootstrap iteration consists in repeatedly generating four frequencies from a multinomial of parameters size = sum(n_ij), i,j = 1, 0 and probabilities (n_11/size, n_10/size, n_10/size, n_00/size). The argument 'xBoot' corresponds to each one of these bootstrap resamples (indiferenly represented in form of a 2x2 "table" or "matrix" or as a numeric vector) In each bootstrap iteration, the value of the "true" known 'dis' is the dissimilarity which was computed from 'x' (a constant, known value in the full iteration) and the values of '^dis' and '^se' are internally computed from the bootstrap data 'xBoot'.

Value

A numeric value, the result of computing (^dis - dis) / ^se.

pablof1988/goSorensen documentation built on July 21, 2023, 8:38 a.m.