Identifying causal effects: Three ways we can fool ourselves with correlations and regression coefficients
Correlation does not equal causality – we all know it! And yet Psychology researchers, especially those working in the field, are often forced to test our causal theories through observational (correlational) designs or experimental designs with observed mediators. Indeed, some of our correlational designs are really very fancy and include long-term longitudinal studies, intensive diary studies, nested or multilevel designs, and complex mediation analysis with bootstrapped indirect effect tests! In this presentation, I will share three ways we can trick ourselves when trying to estimate causal effects: type 1 error, endogeneity, and collider bias. I present all of this from the perspective of a non-expert on causal inference, who stumbled on this literature quite by accident.
Speaker bio: I completed my Master of Industrial and Organisational Psychology and PhD at UWA and am now an Associate Professor at the Future of Work Institute at Curtin University and a Registered Psychologist, endorsed in Organisational Psychology. My research interests lie in personnel recruitment, assessment, and selection in both paid and volunteer settings, hence my interest in personality and other individual differences. I also study how these processes are unfolding in light of new technological and analytical developments, and what the implications are for the applicants.