Surveillance of partially observable systems

Title: Surveillance of partially observable systems
Authors: Mir, Amir
Date: 2008
Abstract: Surveillance of a partially observable system is complex. There are many systems that can be considered as partially observable due to their unknown or partially known structures or the nature of their unknown products and/or partially known results. The impacts of the consumption of genetically modified food (GM) are an example of a system that is only partially observable. The safety of genetically modified foods (GM) products has caused much controversy. Absence of sufficient and reliable information prevents neither certain confidence about the harmlessness of product consumption, nor any certain conclusion to merit a ban for fear of harm. The lack of any reliable or conclusive post-market observation and consumption effects information, make it difficult to establish a global protocol for such products. This paper introduces a model for the analysis of partially observable information from the surveillance of post-market consumption of systems such as genetically modified foods (GM) products. This model uses Markov Chains, paired with a Bayesian updating function to estimate the statistical impacts of surveillance observations and modified surveillance policies. A case study on population health status is used as an illustrative example, which is modeled to demonstrate the impact of policy interventions on simulated data. A cost decision analysis model is also applied to illustrate the impact of policy intervention costs. The model uses a first order Markov chain to estimate the period-over-period change in health status and a Bayesian updating procedure to estimate the population health status based on observations from post-market surveillance. The results show how observation samples can be used to provide information on system changes and improvements.
CollectionTh├Ęses, 1910 - 2010 // Theses, 1910 - 2010
MR48487.PDF14.28 MBAdobe PDFOpen