Premium
Inference of median difference based on the Box–Cox model in randomized clinical trials
Author(s) -
Maruo K.,
Isogawa N.,
Gosho M.
Publication year - 2015
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.6408
Subject(s) - type i and type ii errors , analysis of covariance , statistics , clinical trial , inference , statistical hypothesis testing , statistical inference , randomized controlled trial , power transform , computer science , covariance , mathematics , medicine , artificial intelligence , consistency (knowledge bases)
In randomized clinical trials, many medical and biological measurements are not normally distributed and are often skewed. The Box–Cox transformation is a powerful procedure for comparing two treatment groups for skewed continuous variables in terms of a statistical test. However, it is difficult to directly estimate and interpret the location difference between the two groups on the original scale of the measurement. We propose a helpful method that infers the difference of the treatment effect on the original scale in a more easily interpretable form. We also provide statistical analysis packages that consistently include an estimate of the treatment effect, covariance adjustments, standard errors, and statistical hypothesis tests. The simulation study that focuses on randomized parallel group clinical trials with two treatment groups indicates that the performance of the proposed method is equivalent to or better than that of the existing non‐parametric approaches in terms of the type‐I error rate and power. We illustrate our method with cluster of differentiation 4 data in an acquired immune deficiency syndrome clinical trial. Copyright © 2015 John Wiley & Sons, Ltd.