A robust Bayesian test for identifying context effects in multiattribute decision-making

Area of Science:

• Cognitive psychology
• Decision science
• Behavioral economics

Background:

• Context effects significantly influence multiattribute decision-making, leading to the development of psychological theories.
• Statistical analysis of these context effects has received less attention compared to theoretical explanations.
• Traditional measures like relative choice share (RST) have limitations in accurately capturing these effects.

Purpose of the Study:

• To address weaknesses in the traditional definition and measurement of relative choice share (RST).
• To propose and validate a more appropriate and robust statistical approach for analyzing context effects.
• To introduce the absolute choice share of the target (AST) as a suitable measure for the attraction effect.

Main Methods:

• Critiqued the existing definition of RST and proposed a new, more appropriate measure.
• Conducted large-scale simulations to demonstrate biases in the traditional RST measure.
• Developed and applied a Bayesian approach for estimating a robust RST.
• Introduced and applied the absolute choice share of the target (AST) measure.
• Re-analyzed data from five published studies (N=738) using both traditional and proposed methods.

Main Results:

• The traditional RST measure can lead to biased inferences in statistical analyses.
• The proposed Bayesian approach provides a robust and accurate estimation of RST.
• The absolute choice share of the target (AST) is identified as the appropriate measure for the attraction effect.
• Applying the robust RST and AST measures yielded qualitatively different results in at least 25% of the analyzed studies.

Conclusions:

• Robust statistical testing is crucial for the reliable development of psychological theories on decision-making.
• The proposed statistical measures (robust RST and AST) offer more accurate insights into context effects.
• Re-evaluating existing data with improved methods reveals significant differences, underscoring the importance of methodological rigor.