Term of Award
Doctor of Public Health in Biostatistics (Dr.P.H.)
Document Type and Release Option
Dissertation (restricted to Georgia Southern)
Copyright Statement / License for Reuse
This work is licensed under a Creative Commons Attribution 4.0 License.
Department of Biostatistics, Epidemiology, and Environmental Health Sciences
Committee Member 1
Committee Member 2
Some common summary measures to quantify the test accuracy include sensitivity, specificity, and predictive values. Positive predictive value (PPV) and Negative Predictive value (NPV) are the terms used to describe the posterior probabilities of the presence or absence of disease conditioning on the test results. Given the importance of predictive values in medical diagnosis and prognosis, and the scarcity of research on joint hypothesis testing on positive and negative predictive values when the test/biomarker is continuous, we propose to use two methods: 1) Intersection-union test (IUT) and 2) Joint test for predictive values. The joint hypothesis is formed as H0: PPV =< PPV0 or NPV =< NPV0 against Ha: PPV > PPV0 and NPV > NPV0. Such a hypothesis is multivariate, one-sided, and restrictive. We compare the proposed joint testing methods with the classical multiple testing methods like Holm’s and Bonferroni's corrections. Simulation results indicate that the joint test's power is marginally higher than the IUT test and the competing multiple testing methods, i.e., Holm's procedure and Bonferroni correction. However, the joint test produced the most powerful results compared to all other testing methods. In addition, we used two Wisconsin Breast Cancer data sets (diagnostic and prognostic, respectively) to illustrate and compare these methods.
Modi, Roshni, "Joint Hypothesis Testing on Predictive Values" (2023). Electronic Theses and Dissertations. 2546.
Research Data and Supplementary Material
Available for download on Friday, April 05, 2024