A generalized least squares regression approach for computing effect sizes in single-case research: application...

Area of Science:

• Behavioral Science
• Psychology
• Research Methodology

Background:

• Single-case designs are crucial for evaluating interventions with small sample sizes.
• Existing methods for calculating effect sizes in single-case designs may not adequately address data autocorrelation.
• Accurate effect size estimation is vital for evidence-based practice and research synthesis.

Purpose of the Study:

• To propose and validate a novel method for deriving effect sizes from single-case designs.
• To address the challenge of autoregressive errors in small-sample time-series data common in single-case research.
• To provide a standardized measure of treatment effect magnitude.

Main Methods:

• The proposed method utilizes Generalized Least Squares (GLS) to model autocorrelation in time-series data.
• Regression parameters are estimated to derive an effect size in standard deviation units.
• The method was applied to two published single-case design examples (withdrawal and multiple-baseline).

Main Results:

• The GLS method successfully modeled autocorrelation and estimated effect sizes for the selected single-case designs.
• The derived effect sizes were evaluated against ten desirable criteria for single-case effect measures.
• Application demonstrated the method's utility in quantifying treatment effects across different design phases.

Conclusions:

• Generalized Least Squares (GLS) offers a valuable statistical support for visual analysis in single-case designs.
• The proposed effect size method is applicable to time-series data with autoregressive errors.
• Recommendations for future research and implications for clinical practice are discussed.