A simple ratio-based approach for power and sample size determination for 2-group comparison using Rasch models.

Area of Science:

• Psychometrics
• Clinical Trial Design
• Statistical Modeling

Background:

• Patient-reported outcomes (PRO) are vital in clinical studies, but their design, particularly sample size justification for Rasch model analysis in two-group comparisons, remains challenging.
• Classical sample size formulas (CLASSIC) often underestimate required sample sizes when Rasch models are intended for analysis.
• A correction factor (RATIO) has been proposed to adjust CLASSIC for Rasch model analyses.

Purpose of the Study:

• To investigate the influence of various study design parameters on the RATIO correction factor.
• To identify the most influential parameters for developing a simplified sample size determination method for Rasch modeling.

Main Methods:

• A Monte Carlo simulation approach was employed, varying parameters such as latent trait variance, group effect, sample size per group, number of items, and item difficulty.
• A linear regression model was fitted to explain the RATIO using these design parameters as covariates.

Main Results:

• The number of items and the variance of the latent trait were identified as the most significant predictors of the RATIO, explaining 99.4% of its variation.
• These findings highlight key factors influencing sample size requirements for Rasch model analyses.

Conclusions:

• Adjusting the classical sample size formula with the proposed RATIO offers a simple and dependable method for sample size computation.
• This approach facilitates robust sample size determination for two-group comparisons of PRO data analyzed with Rasch models.