Power analysis on the time effect for the longitudinal Rasch model

Area of Science:

• Biomedical statistics
• Psychometrics
• Health outcomes research

Background:

• Patient-reported outcomes (PROs) are vital in clinical trials but challenging to analyze due to their subjective nature.
• Item response theory (IRT) models, like the Rasch model, are suitable for analyzing PRO data from questionnaires.
• Statistical power analysis is essential for robust study design in social, behavioral, and biomedical sciences.

Purpose of the Study:

• To propose and evaluate a method for assessing statistical power in longitudinal Rasch models with two time points.
• To compare the proposed power evaluation approach with simulation study results.
• To demonstrate the application of the method using a subscale from the SF-36 questionnaire.

Main Methods:

• Development of a novel approach to calculate statistical power for the time effect in a two-timepoint longitudinal Rasch model.
• Comparative analysis of the proposed method against results from a simulation study.
• Application and illustration of the method on empirical data from the SF-36 questionnaire.

Main Results:

• The proposed statistical power evaluation method provides a viable approach for longitudinal Rasch models.
• The performance of the proposed method is comparable to simulation-based evaluations.
• The approach is practical for analyzing time effects in PRO data.

Conclusions:

• The developed method offers a valuable tool for researchers designing clinical trials that utilize longitudinal PRO data analyzed with Rasch models.
• Accurate power analysis enhances the reliability and interpretability of findings related to changes over time in health outcomes.
• This work contributes to the rigorous statistical analysis of subjective health measures in research settings.