Electronic Theses and Dissertations

Date

2024

Document Type

Thesis

Degree Name

Master of Science

Department

Mathematical Sciences

Committee Chair

Ebenezer George

Committee Member

Baris Kopruluoglu

Committee Member

Majid Noroozi

Abstract

This thesis introduces a novel sequential approach for calculating Bayesian P-values, addressing limitations of traditional frequentist methods. Using Monte Carlo simulations, we investigate the properties of sequentially calculated Bayesian P-values under various conditions, comparing them to frequentist counterparts. Our findings demonstrate advantages including incorporation of prior knowledge, continuous evidence measurement, and adaptability to accumulating data. We explore computational challenges, propose optimized algorithms, and address the impact of prior specification, providing guidelines for choosing appropriate priors. This research contributes to the ongoing discussion about P-values in scientific inference and offers a practical framework for researchers adopting Bayesian methods in sequential analyses. The proposed approach has potential applications across various disciplines, particularly where data is collected sequentially, or early decision-making is crucial. Our work bridges the gap between frequentist and Bayesian methods, offering a more nuanced tool for statistical inference.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to ProQuest.

Notes

Open Access

Download

Share

COinS