Browsing by Author "Sujit Ghosh, Committee Member"

Now showing 1 - 13 of 13

Assessing the Effects of Variability in Interest Rate Derivative Pricing
(2007-11-22) Crotty, Michael Thomas; Denis Pelletier, Committee Member; Sujit Ghosh, Committee Member; David Dickey, Committee Member; Peter Bloomfield, Committee Chair
Branch Current State Estimation Method for Power Distribution Systems
(2009-08-09) Jung, Jae Sung; Mesut E. Baran, Committee Chair; Winser E. Alexander, Committee Member; Subhashish Bhattacharya, Committee Member; Sujit Ghosh, Committee Member
Effective management of distribution systems requires analysis tools that can estimate the state of the system (the operating condition). This thesis aims at development of new analysis tools for this purpose. The main tool is the state estimator that will use historical data and the real-time data to estimate the state of the system determined by voltage at all of the nodes of a distribution feeder. This thesis considers the incorporation of voltage measurements in a branch-current-based state estimation (BCSE) program. Original BCSE is designed to include only power and current measurements. The motivation for enhancing BCSE is that with the adoption of large scale automated meter infrastructure (AMI) technologies, voltage measurements will be available at the distribution level. Hence, including these measurements has the potential to improve the accuracy of state estimation. Furthermore, this thesis presents a statistical technique for assessing the BCSE performance. For statistical analysis, 300 Monte Carlo simulations are performed. The overall performance including bias, consistency and quality of estimates is evaluated in order to see the effectiveness of the BCSE method. These concepts of statistical technique are illustrated and tested in this thesis. Finally, since correct connectivity is critical in system operations, topology estimation is expected to become a standard Energy Management System (EMS) function. Hence, two types algorithm are presented for detection and identification of topology error in BCSE. The first approach uses the idea that when the switch status changes, it will affect the measurements. The second approach is based on changing the on/off status of branches one after the other and performing a state estimation in each case. The effectiveness of the proposed approaches is demonstrated. In addition, topology detection results obtained by the two proposed methods are also compared. For testing the revised BCSE, a reduced version of the IEEE 34 node radial test feeder is used. The simulation platform used in this study is developed using C language on Microsoft Visual Studio .NET 2003.
Environmental Health: Evidence at Three Scales.
(2010-04-09) Sanglimsuwan, Karnjana; Erin Sills, Committee Chair; Subhrendu Pattanayak, Committee Chair; Sujit Ghosh, Committee Member; Daniel Robison, Committee Member; Frederick Cubbage, Committee Member
Essays on Multivariate Stochastic Volatility Models Using Wishart Processes: A General Discussion and Dimension Reduction by Latent Factor Structures.
(2010-08-16) Ku, Yu-Cheng; Peter Bloomfield, Committee Chair; Sujit Ghosh, Committee Member; Brian Reich, Committee Member; Min Kang, Committee Member; Ronald Gallant, Committee Member
Evolutionary Computation based Pilot Point Methods for Subsurface Characterization.
(2008-08-11) Jung, Yong; Sujit Ghosh, Committee Member; Sankar Arumugam, Committee Member; Ranji Ranjithan, Committee Co-Chair; G (Kumar) Mahinthakumar, Committee Chair
Hierarchical Bayesian application to instantaneous rates tag-return models
(2009-10-06) Krachey, Matthew James; Ken Pollock, Committee Chair; Kevin Gross, Committee Member; Sujit Ghosh, Committee Member; Joseph Hightower, Committee Co-Chair
Natural mortality has always been a challenging quantity to estimate in harvested populations. The most common approaches to estimation include a regression model based on life history parameters and more recently tag-return models. In recent years, Bayesian methods have been increasingly implemented in ecological models due to their ability to handle increased model complexity and auxiliary datasets. In this dissertation, I explore the implementation of Bayesian methods to analyze tag-return data focusing on natural mortality. Chapter 1 is focused on the addition of two components to the tag-return model framework: random effects and auxiliary data. Auxiliary information on the instantaneous rate of natural mortality is provided through Hoenig's equation relating lifespan to natural mortality, and also implemented through a hierarchical prior. A simulation study validates the performance of the model while an analysis of the classic Cayuga Lake trout dataset demonstrates its use. Chapter 2 adds a change-point allowing for the estimation of two levels of natural mortality and the timing of the discrete-time shift in mortality. Analysis is focused on a Chesapeake Bay striped bass tagging dataset of fish tagged at six years of age and older from 1991-2002. Results show the ability to account for shift in timing. Contrasting with Jiang et al.'s study on the same striped bass dataset, the timing of the change-point was different between the two studies, likely because the Jiang study assumed a fixed tag-reporting probability of 0.43 whereas estimates seem to indicate it may be closer to 0.3. Chapter 3 introduces a change-point allowing for a shift in the tag-reporting probability while assuming a constant natural mortality rate. High reward tags are included in a subset of the data time-series to improve estimation. A factorial simulation design was used to investigate the model performance with different reporting rate and high reward tag scenarios. In general, the model performed very well with little bias except in the case of no high-reward tags. The model performed surprisingly well in a six year study. The results suggest the importance of high reporting rates and/ or auxiliary data sources such as high reward tags.
Issues Regarding Price Risk in Agricultural Commodity Markets.
(2010-06-29) Tejeda, Hernan; Barry Goodwin, Committee Chair; Sujit Ghosh, Committee Member; Denis Pelletier, Committee Member; Nicholas Piggott, Committee Member
Latent Group-Based Interaction Effects in Unreplicated Factorial Experiments.
(2010-09-03) Franck, Christopher; Jason Osborne, Committee Chair; Sujit Ghosh, Committee Member; Jeffrey Thompson, Committee Member; David Dickey, Committee Member
Modeling of Human Exposure to Fine Particulate Matter Using a Stochastic Scenario-Based Model.
(2010-07-08) Cao, Ye; Henry Frey, Committee Chair; Earl Brill, Committee Member; Sujit Ghosh, Committee Member
Semiparametric approaches to inference in joint models for longitudinal and time-to-event data
(2002-05-24) Song, Xiao; Marie Davidian, Committee Co-Chair; Anastasios A. Tsiatis, Committee Co-Chair; Daowen Zhang, Committee Member; Sujit Ghosh, Committee Member; Charles E. Smith, Committee Member
In many longitudinal studies, it is of interest to characterize the relationship between a time-to-event (e.g. survival) and time-dependent and time-independent covariates. Time-dependent covariates are generally observed intermittently and with error. For a single time-dependent covariate, a popular approach is to assume a joint longitudinal data-survival model, where the time-dependent covariate follows a linear mixed effects model and the hazard of failure depends on random effects and time-independent covariates via a proportional hazards relationship. Interest may focus on inference on the longitudinal data process, which is informatively censored by death or withdrawal, or on the hazard relationship. Several methods for fitting such models have been proposed, including regression calibration and likelihood or Bayesian methods. However, most approaches require a parametric distributional assumption (normality) on the random effects. In addition, generalization to more than one time-dependent covariate may become prohibitive. For a single time-dependent covariate, Tsiatis and Davidian (2001) have proposed an approach that is easily implemented and does not require an assumption on the distribution of the random effects. We extend this technique to multiple, possibly correlated,time-dependent covariates. This approach is easy to compute. However, the conditional score approach might be less efficient relative to the likelihood approaches. In addition, inference on the longitudinal data process is not available. To improve the efficiency and meanwhile obtain an estimator for the random effects distribution, we propose to approximate the random effects distribution by the seminonparametric (SNP) densities of Gallant and Nychka (1987), which requires only the assumption that the random effects have a "smooth" density, and take a semiparametric likelihood approach. The EM algorithm is used for implementation. We demonstrate the approaches via simulations and apply them to data from an HIV clinical trial.
Statistical Methods for the Analysis of Forensic DNA Mixtures
(2006-07-11) Beecham, Gary Wayne Jr.; Sujit Ghosh, Committee Member; Dahlia Nielsen, Committee Member; Bruce S. Weir, Committee Chair; Gene Eisen, Committee Member
Forensic DNA mixtures are often interpreted statistically using a likelihood ratio. These ratios are of the form, "The evidence is LR times more likely when assuming the prosecution's hypothesis than when assuming the defense hypothesis." The likelihood ratio calculations rest on the allelic frequencies, yet these frequencies are estimated from only a small portion of the population. Therefore, because of sampling error, the likelihood ratio is an estimate, a random variable. In Chapter 2 the use of a confidence interval to report the variation of likelihood ratios is proposed. The formula for the confidence interval is herein explained and a computer program has been made available. In Chapter 3, a maximum likelihood method is given for the inclusion of peak intensities in forensic DNA mixture likelihood ratio calculations. Observed peak intensities are the result of the underlying composition of the mixture: the amount contributed, and the genotypes of the contributors. This chapter proposes the use of the maximum likelihood method to weight each possible genotype combination by the likelihood of the genotype given the peak intensities. Models based on the Normal and Dirichlet distributions are described. Both models tend to weight more correct genotypes higher, though the Normal model puts much more emphasis on the best model(s) than the Dirichlet. This method can also be applied to certain cases of allele drop out. In the final chapter, several different situations are explored. Four standard cases are considered: single-contributor evidence, two-contributor evidence, the paternity index, and the consideration of relationship by pedigree. These four standard cases are used as an introduction to basic concepts, which are in turn used to discuss more complicated cases later in the chapter. The more complicated cases discussed include analysis of a paternity index from a mixture, relatives and mixtures, consideration of relatives in the presence of population substructure, and a case of canine parentage under varying degrees of relatedness.
Super Population Capture-Recapture Model Augmented with Genetic Data
(2009-11-20) Wen, Zhi; Kenneth H. Pollock, Committee Chair; Charlie Smith, Committee Co-Chair; Sujit Ghosh, Committee Member; James Nichols, Committee Member; Thomas Reiland, Committee Member
Ecologists applying capture-recapture models to animal populations sometimes have access to addition information about individuals' populations of origin. For example, tests that assign an individual's genotype to its most likely source population are increas- ingly used. Here we show how to augment a super population capture-recapture model with such information. We consider a single super population model without age structure, and split the entry probability into separate components due to births in situ and immigration. We show that it is possible to estimate these two probabilities separately. We first consider the case of perfect information about population of origin, where we can distinguish individ- uals born in situ from immigrants with certainty. Then we consider the more realistic case of imperfect information, where we use genetic or other information to assign probabilities to each individual's origin in situ or outside the population. We use a resampling approach to impute the perfect origination assignment data based on the imperfect assignment tests. The integration of data on population of origin with capture-recapture data allows us to determine the contributions of immigration and in situ reproeuction to the growth of the population, an issue of importance to ecologists. Further, the augmentation of capture- recapture data with origination data should improve the preciesion of parameter estimates. We illustrate our new models with capture-recapture and genetic assignment test data from a population of banner-tailed kangaroo rats Dipodomys spectabilis in Arizona. In chapter 4, we evaluate the value of marine reserves for fisheries using tag-return, tag-recapture and telemetry models. We estimate the patch-specific fishing mortality, natu- ral mortality, and movement rates. We first focus on tag-return models for a two-site model with one area a marine reserve and one area a fishing area. We consider tag-return, tag- recapture and telemetry models in various combinations for two site models where one area is a marine reserve and one is subject to regular fishing. Then we illustrate our methods with a comprehensive simulation study.
Variable Selection via Confidence Regions.
(2010-07-20) Gunes, Funda; Howard Bondell, Committee Chair; Leonard Stefanski, Committee Member; Hao Zhang, Committee Member; Sujit Ghosh, Committee Member