The Fred Hutch Biostatistics Program hosts seminars featuring presentations by Hutch and outside scientists to share their latest developments and recent research. Each seminar includes an hour-long presentation and discussion during which speakers showcase their work and findings.
This seminar will be held on Zoom due to the COVID-19 pandemic.
Biostatistics Seminar Series:
“Prediction performance-based testing approach for a global null hypothesis”
Testing a global association between a large set of biomarker measurements and a binary outcome of disease is an important task in biomedical studies, and likelihood-based testing methods have often been the primary approaches for the task. In this talk, we propose a prediction performance-based hypothesis testing approach that combines permutation test using cross validated area under the receiver operating characteristic curve and machine learning techniques for detecting the global association. We particularly focus on introducing the proposed test under two-phase sampling designs, which are very common in biomedical studies, by dealing with an issue arose from inverse sampling probability weights. Furthermore, we develop the use of stacking random forest and a dominant biomarker to achieve the enhanced testing power in both linear and nonlinear problems. Simulation studies show that our proposed test has nominal type 1 error rate and outperforms several likelihood-based tests in terms of power. Illustration with an immunologic marker dataset from an HIV vaccine efficacy trial is also provided.