Comparing Treatment Effects for the AB : BA Crossover Design with Continuous Responses: An Alternative Nonparametric Approach

This work focuses on the nonparametric procedures which are devised as an alternative to analyse the repeated measurements arising in the basic two-treatment, two-period crossover design for continuous responses. Comparing direct effect difference of two treatments is not enough to check for the int...

Full description

Saved in:
Bibliographic Details
Published inJournal of statistical theory and practice Vol. 18; no. 3
Main Authors Bhattacharya, Rahul, Bandyopadhyay, Uttam, Sinha, Abhik
Format Journal Article
LanguageEnglish
Published Cham Springer International Publishing 01.09.2024
Subjects
Mathematics and Statistics
Original Article
Probability Theory and Stochastic Processes
Statistical Theory and Methods
Statistics
Direct treatment effect
Inter-subject evaluation
Asymptotically distribution free
statistic
Step-down approach
Clinical trials
Carryover effect
Crossover design
Asymptotic distribution
Online AccessGet full text

Cover

More Information
Summary:This work focuses on the nonparametric procedures which are devised as an alternative to analyse the repeated measurements arising in the basic two-treatment, two-period crossover design for continuous responses. Comparing direct effect difference of two treatments is not enough to check for the interchangeability of the treatments in a crossover study. Starting with unknown continuous response distributions, nonparametric effect measures for capturing the difference between the direct effects and also that between the carryover effects of two treatments under study are devised using appropriate reliability functional. A few nonparametric vectors are suggested representing actual treatment difference. Mann–Whitney type U -statistics are introduced along with relevant asymptotic results to test for the possible equality of two treatments through inter-subject evaluation. Proposed nonparametric test procedures are compared among themselves and also with competing parametric tests in terms of simulated type-I error probability and power for fixed alternatives. An approximate solution to the problem of sample size determination so that each proposed test achieves a power is also provided. The proposed procedures are further illustrated using a real-life clinical trial data.
ISSN:1559-8608
1559-8616
DOI:10.1007/s42519-024-00386-3