clusterProfile: Calculate biases of numerical data between clusterings.

In raim/segmenTools: Tools for Functional Genome Exploration

Calculate biases of numerical data between clusterings.

Description

Calculate t-tests or wilcoxon (rank sum) tests for each cluster in a clustering against the total distributions. The resulting tables can be plotted with plotOverlaps and sorted along one axis by signficance with plotOverlaps.

Usage

clusterProfile(x, cls, test = stats::t.test, min.obs = 5, replace = FALSE)

Arguments

x	matrix of numeric values where columns are different data sets and rows must correspond to the clustering in argument cls
cls	a clustering of the rows in argument x or a logical TRUE/FALSE table (matrix) with cluster labels as column names and rows corresponding to the rows in argument x
test	test function to be applied, default is t.test. This can be any function that takes the cluster subset of x[cls=<cl>,] as first and the total distribution x as second argument, and returns an object with items statistic and p.value. Negative values in statistic will be differently colored in plotOverlaps and the sign copied to p.value. This package provides w.test, a wrapper for wilcox.test that normalizes the tests' U-statistics such that it can be used in the same way as the t-statistic.
min.obs	minimal number of non-NA observations
replace	test with replacement, i.e., each cluster is tested against the whole data set, including the cluster items ## TODO: instead of passing function, pass a type and handle numbers ## betterer, eg. normalized U-statistic for wilcox - OR handle this ## in plotOverlaps by statistic type!

raim/segmenTools documentation built on March 29, 2025, 9:36 p.m.