plot_corr: Plots to explore the correlations between features

In brightchan/cjbmisc:

Description

Plot the correlations focusing on a variable x vs all the rest of the variables. The workflow is: 1. remove small groups if "min.group.size" is defined; 2. calculate the p values for all pairs of variables 3. select the ones that pass pvalue threshold for plotting. Pvalues are by default non-parametric. Can choose if p.adjust should be used. 4. calculate padj and save the plots and pvalue tables.

Usage

plot_corr_one(
  plotdf,
  x.coln,
  y.coln,
  cat.num.test = "kruskal.test",
  num.num.test = "spearman",
  plot.it = F,
  plot.nrow = NULL,
  plot.ncol = NULL,
  signif.cutoff = 0.05,
  plot.stattest = "np",
  p.adj.method = NULL,
  plot.signif.only = F,
  plot.max = 40,
  min.group.size.x = 3,
  min.group.size.y = 3,
  seed = 999,
  outpdir = NULL,
  plot.w = 7,
  plot.h = 7.5,
  fn.suffix = "",
  ...
)

plot_corr(
  plotdf,
  x.coln,
  y.coln = NULL,
  cat.num.test = "kruskal.test",
  cat.cat.test = "both",
  num.num.test = "spearman",
  plot.it = F,
  plot.nrow = NULL,
  plot.ncol = NULL,
  signif.cutoff = 0.05,
  plot.stattest = "np",
  plot.signif.only = F,
  p.adj.method.each = NULL,
  p.adj.method.all = "bonferroni",
  plot.max = 50,
  seed = 999,
  outpdir = NULL,
  plot.w = 7,
  plot.h = 7.5,
  min.group.size.x = 3,
  min.group.size.y = 3,
  fn.suffix = "",
  ...
)

Arguments

plotdf	dataframe with rows of samples and columns of features.
x.coln	column name of the x axis of the plot
y.coln	character vector of the column names of features to be plotted as y axis
cat.num.test	the significance test to be used for categorical vs numerical variables. Use the name of the r basic tests (Default "kruskal.test").
num.num.test	the significance test to be used for numerical vs numerical variables. Should be "spearman"(Default), "pearson", "kendall", or "lm"(using the pvalue of the independent variable in lm).
plot.it	Whether to plot it out (T/F)
plot.nrow, plot.ncol	The number of rows and columns in the combined plot
plot.stattest	Pass to the "type" parameter in ggbarstats , ggbetweenstats ,ggscatterstats, defining the stats test to be used. Default "np" is non-parametric.
plot.signif.only	whether to plot only the significant items
plot.max	maximum how many plots to be plotted. If set to NULL then plot all.
min.group.size.x	for categorical x, remove groups that are smaller than this number
min.group.size.y	for categorical y, remove groups that are smaller than this number
outpdir	If not NULL, save all plots and pvalues (as table) to the outpdir
plot.w, plot.h	Width and height of each individual plot
fn.suffix	filename suffix
...	pass to ggbarstats , ggbetweenstats ,ggscatterstats
cat.cat.test	the significance test to be used for categorical vs categorical variables. Should be "fisher","chi" or "both"(Default)
padj.method, padj.method.each, padj.method.all	pvalue adjustment method. Should follow ggbetweenstats. In plot_corr_one, if padj.method is specified, pvalue adjustment will be done and only those pass the padj threshold will be plotted. In plot_corr, padj.method.each will be passed to plot_corr_one while padj.method.all will be used to genearate the final padj table, adjusting for all pvalues.

Value

For plot_corr, List of returns from plot_corr_one. For plot_corr_one: List of two: "plot" of ggarrange object which arrange all plot into one, and "pvalues" of a named vector.

brightchan/cjbmisc documentation built on Nov. 5, 2021, 4:12 p.m.