In axzinker/axBoost: axBoost: A collection of useful functions

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(axBoost)

The package axBoost contains a collection of rather unrelated functions, which I found useful to have them exported and available as an own package. At the moment of writing, the functions mainly help to visualize correlations between many variables in a large data set (~ 300 variables). For example, this includes functions to plot a correlogram or to plot double dissocitations (e.g., Asendorpf, Banse, and Mücke, 2002) with and without moderator effects. The functions generating a plot are described below in this vignette, the functionality of functions without graphical output is explained in the example part of the help page of the function.

Asendorpf, J. B., Banse, R., & Mücke, D. (2002). Double dissociation between implicit and explicit personality self-concept: The case of shy behavior. Journal of Personality and Social Psychology, 83(2), 380–393.

The function axCorrgram()

A correlogram of a matrix / data frame is plotted. In the upper right triangle the size of correlations are depicted. Red and blue indicate that the correlations are significant, either positive (red) or negative (blue), also marked with a "*". Marginal significant correlations are colored with darkred (positive) and darkblue (negative), additionally marked with a "+". This is helpful to detect patterns and structures in the data. The diagnoal contains the variable names with a histogram. The lower left triangle depicts a scatter plot with a lowess line. This helps for searching for potential outliers. Up to 16 variables can be correlated in one correlogram. The option fontsize is used to adapt the size of the correlations in the upper right to the size of the boxes.

axCorrgram(axBoost::df1, fontsize = .7)

The function modRegGraph()

This Function plots a moderated regression interaction diagramm and returns the coordinates of the points pred_H_mod_H pred_H_mod_L pred_L_mod_H pred_L_mod_L to be used elswhere (e.g., in another plot). H denotes high (+1 SD), L denotes low (-1 SD).

reg1 <- lm(AutoBehav ~ IndMeasure * mod, data = df1)
summary(reg1)

modRegGraph(coef(reg1), pred = "IndMeasure", mod = "mod", crit = "AutoBehav", plotRange = 1.7)

The function drawDD()

This function draws a double dissociation. This means that two regressions have to be computed. Because this is very prone to error when done by hand, the best way to use this function is in a loop, with predefined predictors, criteria and moderators. The figure below shows a double dissociation with a moderator.

The function is quite versatile and allows to

include moderators (optionally on both paths of a regression)
include covariates
compute robust regressions (using the robust and WRS2 packages)
use the residuals of the covariates as criteria
optionally print the output of the lm() and cor() functions

drawDD(data = axBoost::df1, predUpper = "IndMeasure", predLower = "DirMeasure", 
  critUpper = "AutoBehav", critLower = "ContrBehav", moderator = "mod", 
  title = "Testplot", printReg = TRUE)