Correcting for collider effects and sample selection bias in psychological research

Area of Science:

• Psychological Research Methods
• Statistical Modeling
• Quantitative Psychology

Background:

• Collider variables, common outcomes of independent and dependent variables, introduce bias in psychological research.
• Bias arises from restricted sample composition or statistical adjustment for colliders, impacting result accuracy and generalizability.
• Collider effects are under-recognized in psychology despite their significant methodological implications.

Purpose of the Study:

• To elucidate the conceptual and mathematical underpinnings of collider effects in psychological research.
• To propose and evaluate a method for correcting collider bias stemming from restrictive sample selection.
• To provide practical tools and guidance for researchers to address collider bias.

Main Methods:

• Review of conceptual and mathematical foundations of collider effects.
• Application of Thorndike's Case III adjustment for correcting collider bias in sample selection.
• Two simulation studies to assess the efficacy of the proposed correction method under varying conditions.

Main Results:

• Thorndike's correction method effectively corrects collider bias, even with extreme sample restriction and small sample sizes (N=100).
• Simulation studies demonstrated the method's ability to approximate population correlations accurately.
• Bias and relative bias were analyzed across diverse parameter conditions.

Conclusions:

• The proposed Thorndike's adjustment offers a viable solution for mitigating collider bias in psychological research.
• The method enhances the accuracy and generalizability of statistical findings affected by sample selection.
• The study provides actionable code and discussion for broader application in complex statistical models.