Pairs Plots

For a few recent projects, I have had to quickly get up to speed on new data. If I’m focusing on a relatively small number of variables–perhaps a subset of a much larger dataset–I have found the pairs plot an incredibly useful tool for learning about new data, and getting a visual sense of the distributions of the variables and their relationships. After experimenting with a few different libraries, I have come to really like pairs.panels in the psych (Revelle, 2025) library. I provide an example of pairs.panels below, but I also sometime use what seems to me to be the quite similar in spirit corrgram (Wright, 2021) library.

For this example, I use the data from my Multilevel Workshop site.

Call Relevant Libraries

library(haven) # read Stata data

library(psych) # useful routines for psych data

Get The Data

mydata <- read_dta("simulated_multilevel_data.dta")

Variable Labels

labelled::look_for(mydata) # look at variables and variable labels

 pos variable            label                          col_type missing values
 1   country             country id                     dbl      0             
 2   HDI                 Human Development Index        dbl      0             
 3   family              family id                      dbl      0             
 4   id                  unique country family id       chr      0             
 5   identity            hypothetical identity group v~ dbl      0             
 6   intervention        recieved intervention          dbl      0             
 7   physical_punishment physical punishment in past w~ dbl      0             
 8   warmth              parental warmth in past week   dbl      0             
 9   outcome             beneficial outcome             dbl      0             

Pairs Plot

Options Options Options

Here I set show.points to FALSE for two reasons: by design, most of the variables in my simulated data are uncorrelated, showing the points seemed to somehow create more confusion than clarity; there are a lot of data points, and for that reason as well, showing the points seemed to create more confusion than clarity. You of course can experiment with setting show.points to TRUE

What Do I See? What Am I Looking For?

In the diagonal panel, I see the distribution of each variable by itself. In the bottom panel, there are are smoother plots of each variable against each other. Here I am looking for lines that are not flat–in part or in whole–because this indicates that one variable has a relationship with another. In the top panel, I see the correlations of each variable with each other, and I am looking for either negative or positive numbers that depart significantly from zero.

It seems in this example, that parental warmth and physical punishment have the biggest association with outcome. In regression models, it turns out that intervention also has a relationship with outcome, but that is harder to see here. It is worth investigating even smaller correlations further!

pairs.panels(mydata, # pairs plot
             show.points = FALSE) # don't show data points

References

Revelle, W. (2025). Psych: Procedures for psychological, psychometric, and personality research. Northwestern University. https://CRAN.R-project.org/package=psych

Wright, K. (2021). Corrgram: Plot a correlogram. https://doi.org/10.32614/CRAN.package.corrgram