Equivalence of the mediation, confounding and suppression effect

Area of Science:

• Statistics
• Epidemiology
• Social Sciences

Background:

• Mediation, confounding, and suppression are crucial concepts in statistical analysis.
• Distinguishing between these effects is vital for accurate interpretation of variable relationships.
• Existing methods often focus on individual effects, necessitating a unified statistical perspective.

Purpose of the Study:

• To elucidate the statistical commonalities and differences among mediation, confounding, and suppression.
• To propose methods for calculating confidence intervals for confounding and suppression effects.
• To provide a framework for understanding these effects through a unified statistical lens.

Main Methods:

• Quantifying effects by measuring the change in the independent-dependent variable relationship upon introducing a third variable.
• Applying statistical estimation techniques for effects and standard errors.
• Developing confidence interval methods for confounding and suppression based on existing mediation effect methodologies.

Main Results:

• Mediation and confounding exhibit identical statistical properties.
• Conceptual grounds are essential for differentiating between mediation and confounding.
• Proposed methods allow for confidence interval estimation for confounding and suppression effects.

Conclusions:

• Statistically, mediation and confounding are indistinguishable.
• Conceptual understanding is paramount for distinguishing these effects.
• The proposed methods offer a unified approach to analyzing and quantifying mediation, confounding, and suppression.