Optimizing control variable selection with algorithms: Parsimony and precision in regression analysis

This research note explores the pivotal role of control variables in any tourism and hospitality research that utilizes regression models in statistical analyses. While theory-driven independent variables offer insight into expected effects, the inclusion of control variables is crucial for mitigating potential confounding factors. In an attempt to strike a balance between model complexity and parsimony, researchers face the challenge of selecting the optimal control variables. To address this issue, the study tests three alternative methods: genetic algorithms, lasso models, and the branch and bound algorithm. Despite their underutilization in tourism research, these methods offer efficient means of selecting control variables, enhancing model precision and interpretation without unnecessarily convoluting the model with irrelevant factors.