Physical Sciences and Mathematics Commons^™

Finite Mixtures Of Mean-Parameterized Conway-Maxwell-Poisson Models, Dongying Zhan

Theses and Dissertations--Statistics

For modeling count data, the Conway-Maxwell-Poisson (CMP) distribution is a popular generalization of the Poisson distribution due to its ability to characterize data over- or under-dispersion. While the classic parameterization of the CMP has been well-studied, its main drawback is that it is does not directly model the mean of the counts. This is mitigated by using a mean-parameterized version of the CMP distribution. In this work, we are concerned with the setting where count data may be comprised of subpopulations, each possibly having varying degrees of data dispersion. Thus, we propose a finite mixture of mean-parameterized CMP distributions. An …

Statistical Intervals For Neural Network And Its Relationship With Generalized Linear Model, Sheng Yuan

Theses and Dissertations--Statistics

Neural networks have experienced widespread adoption and have become integral in cutting-edge domains like computer vision, natural language processing, and various contemporary fields. However, addressing the statistical aspects of neural networks has been a persistent challenge, with limited satisfactory results. In my research, I focused on exploring statistical intervals applied to neural networks, specifically confidence intervals and tolerance intervals. I employed variance estimation methods, such as direct estimation and resampling, to assess neural networks and their performance under outlier scenarios. Remarkably, when outliers were present, the resampling method with infinitesimal jackknife estimation yielded confidence intervals that closely aligned with nominal …

High Dimensional Data Analysis: Variable Screening And Inference, Lei Fang

Theses and Dissertations--Statistics

This dissertation focuses on the problem of high dimensional data analysis, which arises in many fields including genomics, finance, and social sciences. In such settings, the number of features or variables is much larger than the number of observations, posing significant challenges to traditional statistical methods.

To address these challenges, this dissertation proposes novel methods for variable screening and inference. The first part of the dissertation focuses on variable screening, which aims to identify a subset of important variables that are strongly associated with the response variable. Specifically, we propose a robust nonparametric screening method to effectively select the predictors …

Deriving The Distributions And Developing Methods Of Inference For R2-Type Measures, With Applications To Big Data Analysis, Gregory S. Hawk

Theses and Dissertations--Statistics

As computing capabilities and cloud-enhanced data sharing has accelerated exponentially in the 21st century, our access to Big Data has revolutionized the way we see data around the world, from healthcare to investments to manufacturing to retail and supply-chain. In many areas of research, however, the cost of obtaining each data point makes more than just a few observations impossible. While machine learning and artificial intelligence (AI) are improving our ability to make predictions from datasets, we need better statistical methods to improve our ability to understand and translate models into meaningful and actionable insights.

A central goal in the …

Beta Mixture And Contaminated Model With Constraints And Application With Micro-Array Data, Ya Qi

Theses and Dissertations--Statistics

This dissertation research is concentrated on the Contaminated Beta(CB) model and its application in micro-array data analysis. Modified Likelihood Ratio Test (MLRT) introduced by [Chen et al., 2001] is used for testing the omnibus null hypothesis of no contamination of Beta(1,1)([Dai and Charnigo, 2008]). We design constraints for two-component CB model, which put the mode toward the left end of the distribution to reflect the abundance of small p-values of micro-array data, to increase the test power. A three-component CB model might be useful when distinguishing high differentially expressed genes and moderate differentially expressed genes. If the null hypothesis above …

Novel Methods For Characterizing Conditional Quantiles In Zero-Inflated Count Regression Models, Xuan Shi

Theses and Dissertations--Statistics

Despite its popularity in diverse disciplines, quantile regression methods are primarily designed for the continuous response setting and cannot be directly applied to the discrete (or count) response setting. There can also be challenges when modeling count responses, such as the presence of excess zero counts, formally known as zero-inflation. To address the aforementioned challenges, we propose a comprehensive model-aware strategy that synthesizes quantile regression methods with estimation of zero-inflated count regression models. Various competing computational routines are examined, while residual analysis and model selection procedures are included to validate our method. The performance of these methods is characterized through …

Estimating And Testing Treatment Effects With Misclassified Multivariate Data, Zi Ye

Theses and Dissertations--Statistics

Clinical trials are often used to assess drug efficacy and safety. Participants are sometimes pre-stratified into different groups by diagnostic tools. However, these diagnostic tools are fallible. The traditional method ignores this problem and assumes the diagnostic devices are perfect. This assumption will lead to inefficient and biased estimators. In this era of personalized medicine and measurement-based care, the issues of bias and efficiency are of paramount importance. Despite the prominence, only few researches evaluated the treatment effect in the presence of misclassifications in some special cases and most others focus on assessing the accuracy of the diagnostic devices. In …

Innovative Statistical Models In Cancer Immunotherapy Trial Design, Jing Wei

Theses and Dissertations--Statistics

A challenge arising in cancer immunotherapy trial design is the presence of non-proportional hazards (NPH) patterns in survival curves. We considered three different NPH patterns caused by delayed treatment effect, cure rate and responder rate of treatment group in this dissertation. These three NPH patterns would violate the proportional hazard model assumption and ignoring any of them in an immunotherapy trial design will result in substantial loss of statistical power.

In this dissertation, four models to deal with NPH patterns are discussed. First, a piecewise proportional hazards model is proposed to incorporate delayed treatment effect into the trial design consideration. …

Novel Nonparametric Testing Approaches For Multivariate Growth Curve Data: Finite-Sample, Resampling And Rank-Based Methods, Ting Zeng

Theses and Dissertations--Statistics

Multivariate growth curve data naturally arise in various fields, for example, biomedical science, public health, agriculture, social science and so on. For data of this type, the classical approach is to conduct multivariate analysis of variance (MANOVA) based on Wilks' Lambda and other multivariate statistics, which require the assumptions of multivariate normality and homogeneity of within-cell covariance matrices. However, data being analyzed nowadays show marked departure from multivariate normal distribution and homoscedasticity. In this dissertation, we investigate nonparametric testing approaches for multivariate growth curve data from three aspects, i.e., finite-sample, resampling and rank-based methods.

The first project proposes an approximate …

Dimension Reduction Techniques In Regression, Pei Wang

Theses and Dissertations--Statistics

Because of the advances of modern technology, the size of the collected data nowadays is larger and the structure is more complex. To deal with such kinds of data, sufficient dimension reduction (SDR) and reduced rank (RR) regression are two powerful tools. This dissertation focuses on these two tools and it is composed of three projects. In the first project, we introduce a new SDR method through a novel approach of feature filter to recover the central mean subspace exhaustively along with a method to determine the dimension, two variable selection methods, and extensions to multivariate response and large p …

Nonparametric Tests Of Lack Of Fit For Multivariate Data, Yan Xu

Theses and Dissertations--Statistics

A common problem in regression analysis (linear or nonlinear) is assessing the lack-of-fit. Existing methods make parametric or semi-parametric assumptions to model the conditional mean or covariance matrices. In this dissertation, we propose fully nonparametric methods that make only additive error assumptions. Our nonparametric approach relies on ideas from nonparametric smoothing to reduce the test of association (lack-of-fit) problem into a nonparametric multivariate analysis of variance. A major problem that arises in this approach is that the key assumptions of independence and constant covariance matrix among the groups will be violated. As a result, the standard asymptotic theory is not …

Measuring Variability In Model Performance Measures, Matthew Rutledge

Theses and Dissertations--Statistics

As data become increasingly available, statisticians are confronted with both larger sample sizes and larger numbers of predictors. While both of these factors are beneficial in building better predictive models and allowing for better inference, models can become difficult to interpret and often include variables of little practical significance. This dissertation provides methods that assist model builders to better understand and select from a collection of candidate models. We study the asymptotic distribution of AIC and propose a graphical tool to assist practitioners in comparing and contrasting candidate models. Real-world examples show how this graphic might be used and a …

Nonparametric Analysis Of Clustered And Multivariate Data, Yue Cui

Theses and Dissertations--Statistics

In this dissertation, we investigate three distinct but interrelated problems for nonparametric analysis of clustered data and multivariate data in pre-post factorial design.

In the first project, we propose a nonparametric approach for one-sample clustered data in pre-post intervention design. In particular, we consider the situation where for some clusters all members are only observed at either pre or post intervention but not both. This type of clustered data is referred to us as partially complete clustered data. Unlike most of its parametric counterparts, we do not assume specific models for data distributions, intra-cluster dependence structure or variability, in effect …

Cancer Phylogenetic Analysis Based On Rna-Seq Data, Tingting Zhai

Theses and Dissertations--Statistics

Studying tumor evolution is a major task to understand the biological mechanism of carcinogenesis, develop new cancer therapies, and prevent drug resistance. We focus on two important questions in tumor evolution. The first question is to quantify intra-tumor heterogeneity, where multiple subclones of tumor cells with distinct transcriptomic profiles. Another question is to estimate the temporal order of alteration of key cancer pathways during tumor evolution. We present a new statistical method to 1) reconstruct the evolutionary history and population frequency of the subclonal lineages of tumor cells and 2) infer temporal order of pathway alterations in tumor evolution for …

Semiparametric And Nonparametric Methods For Comparing Biomarker Levels Between Groups, Yuntong Li

Theses and Dissertations--Statistics

Comparing the distribution of biomarker measurements between two groups under either an unpaired or paired design is a common goal in many biomarker studies. However, analyzing biomarker data is sometimes challenging because the data may not be normally distributed and contain a large fraction of zero values or missing values. Although several statistical methods have been proposed, they either require data normality assumption, or are inefficient. We proposed a novel two-part semiparametric method for data under an unpaired setting and a nonparametric method for data under a paired setting. The semiparametric method considers a two-part model, a logistic regression for …

Bayesian Kinetic Modeling For Tracer-Based Metabolomic Data, Xu Zhang

Theses and Dissertations--Statistics

Kinetic modeling of the time dependence of metabolite concentrations including the unstable isotope labeled species is an important approach to simulate metabolic pathway dynamics. It is also essential for quantitative metabolic flux analysis using tracer data. However, as the metabolic networks are complex including extensive compartmentation and interconnections, the parameter estimation for enzymes that catalyze individual reactions needed for kinetic modeling is challenging. As the pa- rameter space is large and multi-dimensional while kinetic data are comparatively sparse, the estimation procedure (especially the point estimation methods) often en- counters multiple local maximum such that standard maximum likelihood methods may yield …

Estimation Of The Treatment Effect With Bayesian Adjustment For Covariates, Li Xu

Theses and Dissertations--Statistics

The Bayesian adjustment for confounding (BAC) is a Bayesian model averaging method to select and adjust for confounding factors when evaluating the average causal effect of an exposure on a certain outcome. We extend the BAC method to time-to-event outcomes. Specifically, the posterior distribution of the exposure effect on a time-to-event outcome is calculated as a weighted average of posterior distributions from a number of candidate proportional hazards models, weighing each model by its ability to adjust for confounding factors. The Bayesian Information Criterion based on the partial likelihood is used to compare different models and approximate the Bayes factor. …

Statistical Intervals For Various Distributions Based On Different Inference Methods, Yixuan Zou

Theses and Dissertations--Statistics

Statistical intervals (e.g., confidence, prediction, or tolerance) are widely used to quantify uncertainty, but complex settings can create challenges to obtain such intervals that possess the desired properties. My thesis will address diverse data settings and approaches that are shown empirically to have good performance. We first introduce a focused treatment on using a single-layer bootstrap calibration to improve the coverage probabilities of two-sided parametric tolerance intervals for non-normal distributions. We then turn to zero-inflated data, which are commonly found in, among other areas, pharmaceutical and quality control applications. However, the inference problem often becomes difficult in the presence of …

Moment Kernels For T-Central Subspace, Weihang Ren

Theses and Dissertations--Statistics

The T-central subspace allows one to perform sufficient dimension reduction for any statistical functional of interest. We propose a general estimator using a third moment kernel to estimate the T-central subspace. In particular, in this dissertation we develop sufficient dimension reduction methods for the central mean subspace via the regression mean function and central subspace via Fourier transform, central quantile subspace via quantile estimator and central expectile subsapce via expectile estima- tor. Theoretical results are established and simulation studies show the advantages of our proposed methods.

Measuring Change: Prediction Of Early Onset Sepsis, Aric Schadler

Theses and Dissertations--Statistics

Sepsis occurs in a patient when an infection enters into the blood stream and spreads throughout the body causing a cascading response from the immune system. Sepsis is one of the leading causes of morbidity and mortality in today’s hospitals. This is despite published and accepted guidelines for timely and appropriate interventions for septic patients. The largest barrier to applying these interventions is the early identification of septic patients. Early identification and treatment leads to better outcomes, shorter lengths of stay, and financial savings for healthcare institutions. In order to increase the lead time in recognizing patients trending towards septicemia …

Simultaneous Tolerance Intervals For Response Surface And Mixture Designs Using The Adjusted Product Set Method, Aisaku Nakamura

Theses and Dissertations--Statistics

Various methods for constructing simultaneous tolerance intervals for regression models have been developed over the years, but all of them can be shown to be conservative. In this thesis, extensive simulations are conducted to evaluate the degree of conservatism with respect to their coverage probabilities. A new strategy to fit simultaneous tolerance intervals on linear models is proposed by modifying an existing method, which we call the adjusted product set (APS) method. The APS method will also be used to construct simultaneous tolerance bands on response surface and mixture designs.

A Flexible Zero-Inflated Poisson Regression Model, Eric S. Roemmele

Theses and Dissertations--Statistics

A practical problem often encountered with observed count data is the presence of excess zeros. Zero-inflation in count data can easily be handled by zero-inflated models, which is a two-component mixture of a point mass at zero and a discrete distribution for the count data. In the presence of predictors, zero-inflated Poisson (ZIP) regression models are, perhaps, the most commonly used. However, the fully parametric ZIP regression model could sometimes be restrictive, especially with respect to the mixing proportions. Taking inspiration from some of the recent literature on semiparametric mixtures of regressions models for flexible mixture modeling, we propose a …

Unsupervised Learning In Phylogenomic Analysis Over The Space Of Phylogenetic Trees, Qiwen Kang

Theses and Dissertations--Statistics

A phylogenetic tree is a tree to represent an evolutionary history between species or other entities. Phylogenomics is a new field intersecting phylogenetics and genomics and it is well-known that we need statistical learning methods to handle and analyze a large amount of data which can be generated relatively cheaply with new technologies. Based on the existing Markov models, we introduce a new method, CURatio, to identify outliers in a given gene data set. This method, intrinsically an unsupervised method, can find outliers from thousands or even more genes. This ability to analyze large amounts of genes (even with missing …

Serial Testing For Detection Of Multilocus Genetic Interactions, Zaid T. Al-Khaledi

Theses and Dissertations--Statistics

A method to detect relationships between disease susceptibility and multilocus genetic interactions is the Multifactor-Dimensionality Reduction (MDR) technique pioneered by Ritchie et al. (2001). Since its introduction, many extensions have been pursued to deal with non-binary outcomes and/or account for multiple interactions simultaneously. Studying the effects of multilocus genetic interactions on continuous traits (blood pressure, weight, etc.) is one case that MDR does not handle. Culverhouse et al. (2004) and Gui et al. (2013) proposed two different methods to analyze such a case. In their research, Gui et al. (2013) introduced the Quantitative Multifactor-Dimensionality Reduction (QMDR) that uses the overall …

A New Independence Measure And Its Applications In High Dimensional Data Analysis, Chenlu Ke

Theses and Dissertations--Statistics

This dissertation has three consecutive topics. First, we propose a novel class of independence measures for testing independence between two random vectors based on the discrepancy between the conditional and the marginal characteristic functions. If one of the variables is categorical, our asymmetric index extends the typical ANOVA to a kernel ANOVA that can test a more general hypothesis of equal distributions among groups. The index is also applicable when both variables are continuous. Second, we develop a sufficient variable selection procedure based on the new measure in a large p small n setting. Our approach incorporates marginal information between …

Transforms In Sufficient Dimension Reduction And Their Applications In High Dimensional Data, Jiaying Weng

Theses and Dissertations--Statistics

The big data era poses great challenges as well as opportunities for researchers to develop efficient statistical approaches to analyze massive data. Sufficient dimension reduction is such an important tool in modern data analysis and has received extensive attention in both academia and industry.

In this dissertation, we introduce inverse regression estimators using Fourier transforms, which is superior to the existing SDR methods in two folds, (1) it avoids the slicing of the response variable, (2) it can be readily extended to solve the high dimensional data problem. For the ultra-high dimensional problem, we investigate both eigenvalue decomposition and minimum …

Composite Nonparametric Tests In High Dimension, Alejandro G. Villasante Tezanos

Theses and Dissertations--Statistics

This dissertation focuses on the problem of making high-dimensional inference for two or more groups. High-dimensional means both the sample size (n) and dimension (p) tend to infinity, possibly at different rates. Classical approaches for group comparisons fail in the high-dimensional situation, in the sense that they have incorrect sizes and low powers. Much has been done in recent years to overcome these problems. However, these recent works make restrictive assumptions in terms of the number of treatments to be compared and/or the distribution of the data. This research aims to (1) propose and investigate refined …

Multifactor Dimensionality Reduction With P Risk Scores Per Person, Ye Li

Theses and Dissertations--Statistics

After reviewing Multifactor Dimensionality Reduction(MDR) and its extensions, an approach to obtain P(larger than 1) risk scores is proposed to predict the continuous outcome for each subject. We study the mean square error(MSE) of dimensionality reduced models fitted with sets of 2 risk scores and investigate the MSE for several special cases of the covariance matrix. A methodology is proposed to select a best set of P risk scores when P is specified a priori. Simulation studies based on true models of different dimensions(larger than 3) demonstrate that the selected set of P(larger than 1) risk scores outperforms the single …

High Dimensional Multivariate Inference Under General Conditions, Xiaoli Kong

Theses and Dissertations--Statistics

In this dissertation, we investigate four distinct and interrelated problems for high-dimensional inference of mean vectors in multi-groups.

The first problem concerned is the profile analysis of high dimensional repeated measures. We introduce new test statistics and derive its asymptotic distribution under normality for equal as well as unequal covariance cases. Our derivations of the asymptotic distributions mimic that of Central Limit Theorem with some important peculiarities addressed with sufficient rigor. We also derive consistent and unbiased estimators of the asymptotic variances for equal and unequal covariance cases respectively.

The second problem considered is the accurate inference for high-dimensional repeated …

Accounting For Matching Uncertainty In Photographic Identification Studies Of Wild Animals, Amanda R. Ellis

Theses and Dissertations--Statistics

I consider statistical modelling of data gathered by photographic identification in mark-recapture studies and propose a new method that incorporates the inherent uncertainty of photographic identification in the estimation of abundance, survival and recruitment. A hierarchical model is proposed which accepts scores assigned to pairs of photographs by pattern recognition algorithms as data and allows for uncertainty in matching photographs based on these scores. The new models incorporate latent capture histories that are treated as unknown random variables informed by the data, contrasting past models having the capture histories being fixed. The methods properly account for uncertainty in the matching …