Sample size estimation in educational intervention trials with subgroup heterogeneity in only one arm

Area of Science:

• Psychology
• Biostatistics
• Health Services Research

Background:

• Psycho-social interventions often involve complex delivery methods.
• Standard sample size calculations may not apply when subgroup heterogeneity is present in only one study arm.
• Clustered randomized trials typically assume heterogeneity across all arms.

Purpose of the Study:

• To develop novel closed-form sample size and power formulas for psycho-social interventions.
• To address the unique challenge of subgroup heterogeneity existing in only the experimental group.
• To optimize study parameters for efficient resource allocation and enhanced statistical power.

Main Methods:

• Derivation of closed-form formulas for sample size and power calculations.
• Consideration of continuous outcomes at single time points and longitudinally.
• Optimization of parameters like subgroup number and time points under cost constraints.

Main Results:

• Novel formulas accounting for one-sided subgroup heterogeneity.
• Methods for comparing continuous outcomes at single and multiple time points.
• Guidance on balancing sample size, power, and measurement constraints.

Conclusions:

• The presented formulas offer a more accurate approach for studies with specific intervention designs.
• Optimized parameter selection can lead to minimized sample sizes and maximized statistical power.
• These methods support more efficient and effective psycho-social intervention research.