Skip to Main Content
Frequently Asked Questions
Submit an ETD
Global Search Box
Need Help?
Keyword Search
Participating Institutions
Advanced Search
School Logo
Files
File List
Thesis.pdf (11.21 MB)
ETD Abstract Container
Abstract Header
Bayesian Models for Computer Model Calibration and Prediction
Author Info
Vaidyanathan, Sivaranjani
Permalink:
http://rave.ohiolink.edu/etdc/view?acc_num=osu1435527468
Abstract Details
Year and Degree
2015, Doctor of Philosophy, Ohio State University, Statistics.
Abstract
Computer models have increasingly been used to study physical processes from which direct observations are difficult or impossible to obtain. The use of computer models in place of observations brings with it a whole new set of questions of interest to researchers. In this thesis, two of these questions are addressed. First, the problem of computer model calibration is considered. Here, a few physical observations are available along with output from runs of the computer model. In the calibration problem, a subset of parameters that affect the response of the process are assumed to have some unknown "true" value for the process being studied. A statistical model using observed data and computer model output is then used to learn about these parameters. Existing approaches to address this problem can be highly time and resource intensive to use in practice when the response of the computer model is of high dimension and there is a limit on the number of runs that can be obtained from the computer model. In this thesis, two new approaches to computer model calibration are presented. The proposed approaches are then applied to a real world application. Results from this model are compared with results from an existing popular approach. The second problem that is addressed in this thesis is in the use of multiple computer models that each model the same physical process but vary in terms of their fidelity to the true process. If physical observations are not available in this setting, an area of interest is in predicting the output of the most accurate model at an arbitrary input given the output of the less accurate models. A statistical approach to solve this problem is proposed. Three different models are presented to illustrate this approach. These models are applied to a real world application and the results are compared. The thesis ends with a discussion on the problems that are addressed and on possible extensions to this methodology.
Committee
L Mark Berliner (Advisor)
William Irwin Notz (Committee Member)
Matthew Timothy Pratola (Committee Member)
Radu Herbei (Committee Member)
Pages
130 p.
Subject Headings
Statistics
Keywords
Bayesian Modeling, Computer Experiments, Calibration, Prediction
Recommended Citations
Refworks
EndNote
RIS
Mendeley
Citations
Vaidyanathan, S. (2015).
Bayesian Models for Computer Model Calibration and Prediction
[Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1435527468
APA Style (7th edition)
Vaidyanathan, Sivaranjani.
Bayesian Models for Computer Model Calibration and Prediction.
2015. Ohio State University, Doctoral dissertation.
OhioLINK Electronic Theses and Dissertations Center
, http://rave.ohiolink.edu/etdc/view?acc_num=osu1435527468.
MLA Style (8th edition)
Vaidyanathan, Sivaranjani. "Bayesian Models for Computer Model Calibration and Prediction." Doctoral dissertation, Ohio State University, 2015. http://rave.ohiolink.edu/etdc/view?acc_num=osu1435527468
Chicago Manual of Style (17th edition)
Abstract Footer
Document number:
osu1435527468
Download Count:
473
Copyright Info
© 2015, all rights reserved.
This open access ETD is published by The Ohio State University and OhioLINK.