Beyond Systematic and Unsystematic Responding: Latent Class Mixture Models to Characterize Response Patterns in...

Area of Science:

• Behavioral Economics
• Psychology
• Public Policy Analysis

Background:

• Operant behavioral economic methods are vital for assessing reinforcer efficacy and public policy.
• Current methods for screening discounting data may exclude valid data or include invalid data due to rigid assumptions.
• Inaccurate data screening impacts study power, precision of estimates, and generalizability of findings.

Purpose of the Study:

• To review existing approaches for characterizing discounting behavior.
• To present and demonstrate a novel, data-driven method, Latent Class Mixed Modeling (LCMM), for analyzing discounting data.
• To offer LCMM as a supplement to existing methods for inspecting and screening discounting data.

Main Methods:

• Review of existing rule-based and statistical approaches for discounting data screening.
• Application of Latent Class Analysis to identify distinct responder groups within discounting data.
• Demonstration of Latent Class Mixed Modeling using a publicly available dataset.

Main Results:

• Latent Class Mixed Modeling effectively characterizes longitudinal choice patterns into distinct classes.
• The approach identifies groups of responders with characteristic differences from the general sample.
• Mixed-effects models are robust for less systematic data series when distinct classes are absent.

Conclusions:

• Latent Class Mixed Modeling offers a data-driven alternative for analyzing discounting data.
• This method enhances the accuracy of data screening in behavioral economic research.
• LCMM can improve the reliability and validity of findings in large-scale studies and policy evaluations.