Bayesian generalized linear mixed effects models for deception detection analyses or: How I learned to stop aggregating veracity judgments and embraced Bayesian Mixed Effects models

Historically, deception detection research has relied on factorial analyses of response accuracy to make inferences. But this practice overlooks important sources of variability resulting in potentially misleading estimates and may conflate response bias with participants’ underlying sensitivity to detect lies from truths. We offer an alternative approach using Bayesian Generalized Linear Mixed Models (BGLMMs) within a Signal Detection Theory (SDT) framework to address these limitations. Our approach incorporates individual differences from both judges and senders, which are a principal source of spurious findings in deception research. By avoiding data transformations and aggregations, this methodology outperforms traditional methods and provides more informative and reliable effect estimates. The proposed framework offers researchers a powerful tool for analyzing deception data and advances our understanding of veracity judgments. All code and data are openly available.