Causal mediation effects in single case experimental designs.

Area of Science:

• Psychology
• Behavioral Science
• Research Methodology

Background:

• Single case experimental designs (SCEDs) are crucial for evaluating treatment effects in various fields.
• Mediation analysis, recently applied to SCEDs, decomposes treatment-outcome effects into direct and indirect components to explore causal mechanisms.
• Causal mediation analysis methodology clarifies essential causal assumptions for mediation analysis.

Purpose of the Study:

• To derive causal mediation effects and standard errors using piecewise linear regression models for mediators and outcomes in SCEDs.
• To evaluate the performance of regression estimators and standard errors for causal mediation analysis.
• To demonstrate that causal direct and indirect effects encompass both level and trend changes, unlike previous methods.

Main Methods:

• Utilized piecewise linear regression models to estimate causal mediation effects and standard errors.
• Conducted a simulation study to compare the performance of different confidence intervals (Monte Carlo vs. normal theory).
• Analyzed both level and trend changes in the context of direct and indirect effects.

Main Results:

• Monte Carlo confidence intervals showed accurate Type I error rates and higher power for causal indirect effects.
• Normal theory confidence intervals demonstrated accurate Type I error rates and higher power for causal direct and total effects.
• The study confirmed that causal mediation effects integrate both level and trend adjustments.

Conclusions:

• The proposed regression-based method provides a robust approach to causal mediation analysis in SCEDs.
• Different confidence interval methods are optimal for indirect versus direct and total effects, respectively.
• This research advances the understanding of causal processes in single-case research.