Premium
A note on testing families of hypotheses using graphical procedures
Author(s) -
Maurer Willi,
Bretz Frank
Publication year - 2014
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.6267
Subject(s) - multiple comparisons problem , graphical model , null hypothesis , statistical hypothesis testing , computer science , false discovery rate , variety (cybernetics) , test (biology) , type i and type ii errors , statistics , machine learning , econometrics , data mining , mathematics , artificial intelligence , paleontology , biochemistry , chemistry , biology , gene
Regulatory guidelines for drug development suggest a strong control of the familywise error rate, when multiple hypotheses are simultaneously tested in confirmatory clinical trials. Accordingly, a variety of multiple test procedures exist for pharmaceutical trial applications, which are typically defined on a single structured family of null hypotheses. For some confirmatory clinical trials with multiple objectives, however, it is advantageous to address the arising multiplicity problems via grouped families of hypotheses. Graphical test procedures have been developed to construct, visualize, and perform multiple test procedures that are tailored to either a single or multiple families of structured hypotheses of interest. This note complements the existing literature by introducing a general algorithm to calculate adjusted p ‐values for any sequentially rejective graphical test procedure where the test procedures within the individual families of hypotheses are not necessarily graphical. Copyright © 2014 John Wiley & Sons, Ltd.