Nonparametric analysis of quality-of-life measures for randomized clustered design: the R package npclust.

Area of Science:

• Biomedical and behavioral sciences
• Statistical methodology
• Health outcomes research

Background:

• Quality-of-Life (QoL) research often uses pre-post intervention designs with complex within-subject correlations.
• Parametric methods are questionable due to unknown data distributions, outliers, and rating scale measurements in QoL data.
• There is a need for robust statistical procedures with minimal assumptions for QoL outcome analysis.

Purpose of the Study:

• To introduce nonparametric methods for analyzing QoL data in clustered randomized designs.
• To address the limitations of parametric and semi-parametric methods in QoL research.
• To provide researchers with user-friendly tools for QoL data analysis.

Main Methods:

• Development and illustration of the R package npclust for nonparametric analysis of QoL data.
• Application of nonparametric methods to handle unknown distributions and complex correlations in QoL outcomes.
• Utilizing a pre-post intervention factorial design with multiple observations.

Main Results:

• The npclust package offers procedures for data pre-processing, hypothesis testing, and confidence interval computation for QoL data.
• Nonparametric methods provide a reliable approach for analyzing QoL outcomes with minimal distributional and correlation assumptions.
• The methods were illustrated using Pediatric Asthma Quality-of-Life data from the ARTIS study.

Conclusions:

• Nonparametric methods are crucial for accurate and reliable analysis of QoL data, especially in clustered randomized designs.
• The npclust R package facilitates accessible and user-friendly application of these advanced statistical techniques.
• This work supports researchers in conducting rigorous QoL research in biomedical and behavioral sciences.