Statistical analysis is used quite heavily in production operations. To use certain advanced statistical approaches such as Bayesian analysis, statistical models must be built. This thesis demonstrates the process of building the Bayesian models and addresses some of the classical limitations by presenting mathematical examples and proofs, by demonstrating the process with experimental and simulated implementations, and by completing basic analysis of the performance of the implemented models. From the analysis, it is shown that the performance of the Bayesian models is directly related to the amount of separation between the likelihood distributions that describe the behavior of the data features used to generate the multivariate Bayesian models. More specifically, the more features that had clear separation between the likelihood distributions for each possible condition, the more accurate the results were. This is shown to be true regardless of the quantity of data used to generate the model distributions during model building. In cases where distribution overlap is present, it is found that models performance become more consistent as the amount of data used to generate the models increases. In cases where distribution overlap is minimal, it is found that models performance become consistent within 4-6 data sets.
| Identifer | oai:union.ndltd.org:GATECH/oai:smartech.gatech.edu:1853/55036 |
| Date | 27 May 2016 |
| Creators | Locks, Stephanie Isabel |
| Contributors | Kurfess, Thomas R. |
| Publisher | Georgia Institute of Technology |
| Source Sets | Georgia Tech Electronic Thesis and Dissertation Archive |
| Language | en_US |
| Detected Language | English |
| Type | Thesis |
| Format | application/pdf |
Page generated in 0.0021 seconds