Visualizing Linear Models: An R Bag of Tricks
Michael Friendly
Oct-Nov, 2022
0: Computing set up
To follow along and run any tutorials or examples, you will need the following:
A recent version of R, download from here.
A recent version of RStudio, download from here.
The following R packages are used:
arm, broom, candisc, car, corrgram, dplyr, effects, ggplot2, glue, heplots, knitr, learnr, MASS, modelsummary, mvinfluence, palmerpenguins, stargazer, rgl, tidyverse, visreg
You can install all required R packages at once by running the script install-vismlm-pkgs.r. Alternativelty, use the following code in the R Studio console or from an editor window.
pkgs <- c("arm", "broom", "candisc", "car", "corrgram", "dplyr", "effects", "ggplot2",
"glue", "heplots", "knitr", "learnr", "MASS", "modelsummary",
"mvinfluence", "palmerpenguins", "stargazer", "rgl", "rpart", "rpart.plot",
"tidyverse", "visreg")
# install the above, along with any dependencies
install.packages(pkgs, dependencies=TRUE)1: Getting started: Linear Models
Topics:
- Why plot your data?
- Data plots
- Effect plots
- Diagnostic plots
Materials:
Lab exercise:
Visualizing regression models: duncan-reg.html || duncan-reg.R
Examples:
- Prestige data: prestige-ex.R || prestige-ex.html
- Penguin data: penguins-lm-ex.R || penguins-lm-ex.html
- Coefficient plots: coefplots-demo.R || coefplots-demo.html
- Diagnostic plots: prestige-diagnose.R || prestige-diagnose.html
Learning more
- Building better tables in R How to make tables people ACTUALY want to read.
- How to make beautiful tables in R
- Visualizing Trends of Multivariate Data in R using ggplot2
- Claus Wilke, SDS 375/395 Data Visualization in R This is a comprehensive course in R graphics (mainly ggplot2 & friends), based on Wilke’s Fundamentals of Data Visualization.
- Effect plots are illustrated in the predictor effects gallery vignette
- Visreg: An R package for the visualization of regression models
- Exploratory multivariate plots of the Palmer Penguins data Penguin data: Multivariate EDA
2: Multivariate Models
Topics:
- Brief review of the MLM
- Data ellipses
- HEplot framework
Materials:
- Lecture notes: 1up PDF || 4up PDF || video
- Bonus lecture: PCA-animation
- Lab exercise:
- Visualizing Multivariate Data: Penguins penguins-mlm.Rmd Open this in R Studio & Run Document.
- Examples:
- Math scores: Simple demo of MLMs mathscore-simple.R
- Math scores: HE plot examples mathscore-ex.R || mathscore-ex.html
- Penguins data: Multivariate EDA vignette
- Diabetes data: heplots and candisc examples vignette
Learning more
- Fox, Friendly & Weisberg (2013), Hypothesis Tests for Multivariate Linear Models Using the car Package http://www.datavis.ca/papers/Fox+Friendly+Weisberg.pdf
- Friendly & Sigal (2017), Graphical methods for multivariate linear models in psychological research: An R tutorial DOI: 10.20982/tqmp.13.1.p020 http://www.tqmp.org/RegularArticles/vol13-1/p020/p020.pdf
- MANOVA and HE plots This vignette illustrates the use of MANOVA, HE plots and canonical discriminant analysis in the analysis of the Palmer penguins data.
- Friendly (2020), HE plot Examples
heplotspackage documentation, http://friendly.github.io/heplots/
3: Examples & Extensions
Topics:
- MANOVA
- Robust MLMs
- MMREG
- Canonical correlations
- MANCOVA and homogeneity of regression
- Homogeneity of covariance matrices
Materials:
- Lecture notes: 1up PDF || 4up PDF || video
- Lab exercise:
- Visualizing MMRA Models rohwer-mmra.html || rohwer-mmra.R
- Examples:
- Neuro Cognitive measures NeuroCog-ex.R || NeuroCog-ex.html
- Social cognition measures SC-ex.R || SC-ex.html
- Rohwer MMRA data Rohwer-MMRA-ex.R || Rohwer-MMRA-ex.html
Learning more
- Friendly & Sigal (2016) Analysis of Rohwer Data with
MANCOVA, https://rpubs.com/friendly/Rohwer-MANCOVA-ex
- Friendly & Sigal (2018) Visualizing Tests for Equality of Covariance Matrices http://www.datavis.ca/papers/EqCov-TAS.pdf
Copyright © 2020 Michael Friendly. All rights reserved.
friendly AT yorku DOT ca
orcid.org/0000-0002-3237-0941