apsl2lme: A Model-Selection Diagnostic Tool for Hierarchical Linear Models

Abstract

A primary goal in regression is to choose the simplest model that provides the best fit to the observed data (Thompson, 2006; West, Welch, & Galecki, 2014). In ordinary least squares regression, this may be a simple process of examining the relationship between the number of predictors and the resulting Multiple R² or Adjusted R² (Thompson, 2006). However, in multilevel modeling, the model selection process is more complicated. Not only must researchers consider which variables to include in the model, they must also determine whether level-1 variables should be modeled as random effects (West et al., 2104). The purpose of this paper is to present a software-based model-selection diagnostic tool that supports two-level models with a single grouping factor.