variablesRelation: Calculate all pairwise variable relations

In moosik/snorm: Supporting function for supervised exploration of data structure

Description

Given a data frame the function will calculate pairwise correlations between all columns. This function is usually used to understand if any of the technical variables in the data are correlated with the biological variables (to avoid loosing the signal during data normalization). The following tests are used: between a factor and a numeric variable or between two numeric variables - anova F test, between two factor variables - Chi-square test. If simulate.p.val = TRUE then the p-value for the Chi-square test is computed for a Monte Carlo test with monte.carlo.reps replicates. We suggest removing any columns ID columns from the data prior to using this function (definition of an ID column: number of levels is the same as the number of rows in the data frame)

Usage

variablesRelation(df, multiple.testing = c("fdr", "BY", "q"))

Arguments

df	data frame for which relationship between columns needs to be calculated
multiple.testing	method to use for multiple testing adjustment, currently supported methods are Benjamini & Hochberg's FDR ("fdr"), Benjamini & Yekutieli ("BY") or Storey's q-value ("q")

Value

plot

ggplot object with the tile plot where two colors are used to highlight significant relationships (alpha is less or equal to 0.05)

table

a data frame with all pairwise tests: variable 1 name, variable 2 name, P value, test statistic, test name, adjusted P value using one of methods

moosik/snorm documentation built on May 23, 2019, 6:11 a.m.