Distributional moderation analysis: Unpacking moderation effects in intervention research.

Area of Science:

• Statistical modeling
• Intervention research
• Educational psychology

Background:

• Traditional moderation and subgroup analyses focus on average intervention effects.
• Intervention effect heterogeneity can manifest beyond group means, affecting variance, skewness, kurtosis, and scale endpoints.
• Existing methods may not fully capture the complexity of intervention effect modifiers.

Purpose of the Study:

• Introduce distributional moderation analysis using inflated Generalized Additive Models for Location, Scale, and Shape (GAMLSS).
• Provide a framework to holistically characterize intervention effect modifiers by modeling various distributional parameters.
• Demonstrate the application of distributional moderation analysis in school-based intervention research.

Main Methods:

• Utilized the framework of inflated Generalized Additive Models for Location, Scale, and Shape (GAMLSS).
• Simultaneously modeled conditional mean-, variance-, skewness-, and kurtosis-based intervention effects.
• Analyzed moderated treatment effects at response scale endpoints (floor/ceiling effects) using data from a randomized controlled trial on classroom management.

Main Results:

• Traditional mean-focused analysis indicated intervention reduced disruptive behavior only for students receiving special education.
• Distributional moderation analysis revealed a general decrease in disruptive behavior for at-risk students.
• Race moderated the average decrease, and special education status influenced the likelihood of exhibiting no disruptive behavior, not the average behavior itself.

Conclusions:

• Distributional moderation analysis provides a more fine-grained characterization of treatment effect modifiers compared to traditional methods.
• This approach offers a valuable complementary tool for understanding complex intervention impacts.
• The study highlights the importance of examining effects beyond the conditional mean to fully understand intervention effectiveness.