Resurrecting the individual in behavioral analysis: Using mixed effects models to address nonsystematic discounting...

Area of Science:

• Behavioral economics
• Decision science
• Psychology

Background:

• Delay discounting (DD) and probability discounting (PD) functions are typically monotonic.
• Some individuals exhibit nonsystematic functions, classified as Type 1 (random choices) or Type 2 (shallow slopes) by Johnson and Bickel (2008).
• Existing algorithms often lead to the removal of ~20% of participants, potentially impacting study power and sample representativeness.

Purpose of the Study:

• To apply mixed effects regression modeling to account for individual differences in DD and PD functions.
• To assess the impact of removing participants with nonsystematic functions.
• To advocate for mixed effects models as an alternative to participant removal in behavioral analysis.

Main Methods:

• Utilized mixed effects regression modeling to analyze DD and PD functions.
• Assessed model estimates for Type 1 and Type 2 nonsystematic functions.
• Compared the properties of samples with and without removed participants.

Main Results:

• Nonsystematic functions (Type 1 and Type 2) showed shallower slopes and increased biases for larger amounts compared to systematic functions.
• Removing participants with nonsystematic functions would significantly alter the final sample's properties.
• Mixed effects models effectively account for between-participant variation using random effects.

Conclusions:

• Mixed effects models provide a robust method for analyzing nonsystematic DD and PD functions without participant removal.
• Participant removal can introduce undesirable bias and reduce statistical power.
• The use of mixed effects models is recommended for future analyses in behavioral science to preserve sample integrity and enhance analytical rigor.