Multivariate nonparametric analysis for the two-period crossover design with application in clinical trials

Area of Science:

• Biostatistics
• Statistical Methods

Background:

• Crossover designs are efficient for comparing treatments within subjects.
• Multivariate responses present unique analytical challenges in crossover trials.
• Traditional parametric methods may fail when assumptions are violated.

Purpose of the Study:

• To present nonparametric methods for analyzing two-treatment, two-period crossover designs with multivariate responses.
• To provide a robust alternative to traditional analyses when assumptions are questionable.
• To address the broader class of hypotheses arising from multivariate responses.

Main Methods:

• Forming within-subject sums and differences.
• Utilizing multivariate analysis of variance for standard tests.
• Developing multivariate nonparametric tests based on ranks.

Main Results:

• Nonparametric tests can be constructed for carry-over and direct treatment effects.
• Rank-based multivariate tests offer a realistic alternative to parametric methods.
• The study formulates nonparametric tests for a wider range of hypotheses in multivariate crossover designs.

Conclusions:

• Nonparametric methods are valuable for multivariate crossover designs, especially when parametric assumptions are not met.
• These methods extend the analysis of crossover trials to complex multivariate hypotheses.
• The proposed techniques enhance the reliability of treatment effect analysis in such designs.