The interpretation and presentation of empirical findings from (generalized) linear models has come a long way in the social sciences. Researchers increasingly visualize substantively meaningful quantities of interest such as expected values, first differences, and average marginal effects and consistently include uncertainty estimates in the form of analytical, simulation-based, or bootstrapped confidence intervals. However, existing interpretations and presentations are typically restricted to bivariate patterns which show (changes in) expected values as function of a single predictor, holding all else constant. This can be a significant limitation, especially when substantive inquiries focus on the interplay of two variables in predicting an outcome. To interpret and visualize such applications effectively, researchers must extend their presentations to include a third dimension. In this Methods Bites Tutorial, Denis Cohen and Nick Baumann introduce and showcase the regplane3D package, a tool for plotting 3D regression predictions in R.
After reading this blog post and engaging with the applied examples, readers will be able to:
generate the quantities of interest from regression models, including expected values over a grid of predictor values and their confidence intervals.
plot these quantities in three-dimensional visualizations using regplane3D.
