New Generalized p-Value Approach for Noninferiority Analysis

The underlying statistical exercise in some noninferiority studies is to test a composite null hypothesis of a nonzero difference in the means of two normally distributed populations. These noninferiority studies are often analyzed under equal variances assumption. Unequal variances frequently occur...

Full description

Saved in:
Bibliographic Details
Published inStatistics in biopharmaceutical research Vol. 8; no. 1; pp. 35 - 39
Main Author Koti, Kallappa M.
Format Journal Article
LanguageEnglish
Published Taylor & Francis 02.01.2016
Subjects
Behrens-Fisher problem
Beta distribution
Noncentrality parameter
Probability value
Sample size
Online AccessGet full text

Cover

More Information
Summary:The underlying statistical exercise in some noninferiority studies is to test a composite null hypothesis of a nonzero difference in the means of two normally distributed populations. These noninferiority studies are often analyzed under equal variances assumption. Unequal variances frequently occur in practice. In this article, we propose a noncentral t-distribution-based generalized test for proving noninferiority in the presence of unequal variances. We provide a simple SAS IML code to calculate the new generalized p-value. We compare the new generalized p-value with that of the Student's central t-distribution-based Tsui-Weerahandi generalized test. We conclude that Tsui-Weerahandi generalized test is conservative.
ISSN:1946-6315
1946-6315
DOI:10.1080/19466315.2015.1093958