Determination of Probabilistic Risk Assessment Initiating Event Frequency Baseline Periods using Change-Point Analysis
David Teolis, Distance Degree Program, M.S. Candidate, Department of Statistics, Colorado State University
Tuesday, March 30, 2010
9:00 a.m., room 008, Statistics Bldg
Change-point analysis is typically performed on count data generated over several time periods to identify if and when changes take place in the underlying frequency of occurrence of a particular event. Many methods are available to conduct change-point analysis to estimate when a process has changed and what the process produces such as the number of occurrences of an event. A change-point model can be viewed as a time series in which the data generating process is dependent on different regimes or states over time. The application of change-point analysis is popular in many disciplines and relatively straightforward to perform for change-point models with one or two changepoints. This report uses the frequency of general transient events at nuclear power plants in the United States with pressurized water reactors (PWRs) to demonstrate the application of several methods for performing change-point analysis. The purpose of the analysis is to determine if a change-point exists and what the frequency of occurrence is after the change-point. The results for each method are compared and discussed.
Dr. Chihoon Lee, Advisor
Dr. Dan Cooley, Committee Member
Dr. Ken Berry, Sociology, Outside Member