Empirical likelihood (EL) is a recently developed nonparametric method of statistical inference. It has been shown by Owen (1988,1990) and many others that empirical likelihood ratio (ELR) method can be used to produce nice confidence intervals or regions. Owen (1988) shows that -2logELR converges to a chi-square distribution with one degree of freedom subject to a linear statistical functional in terms of distribution functions. However, a generalization of Owen's result to the right censored data setting is difficult since no explicit maximization can be obtained under constraint in terms of distribution functions. Pan and Zhou (2002), instead, study the EL with right censored data using a linear statistical functional constraint in terms of cumulative hazard functions. In this dissertation, we extend Owen's (1988) and Pan and Zhou's (2002) results subject to non-linear but Hadamard differentiable statistical functional constraints. In this purpose, a study of differentiable functional with respect to hazard functions is done. We also generalize our results to two sample problems. Stochastic process and martingale theories will be applied to prove the theorems. The confidence intervals based on EL method are compared with other available methods. Real data analysis and simulations are used to illustrate our proposed theorem with an application to the Gini's absolute mean difference.

Medical cost analysis is an important part of treatment evaluation. Since resources are limited in society, it is important new treatments are developed with proper costconsiderations. The mean has been mostly accepted as a measure of the medical cost analysis. However, it is well known that cost data is highly skewed and the mean could be highly influenced by outliers. Therefore, in many situations the mean cost alone cannot offer complete information about medical costs. The quantiles (e.g., the first quartile, median and third quartile) of medical costs could better represent the typical costs paid by a group of individuals, and could provide additional information beyond mean cost. For a specified patient population, cost estimates are generally determined from the beginning of treatments until death or end of the study period. A number of statistical methods have been proposed to estimate medical cost. Since medical cost data are skewed to the right, normal approximation based confidence intervals can have much lower coverage probability than the desired nominal level when the cost data are moderately or severely skewed. Additionally, we note that the variance estimators of the cost estimates are analytically complicated. In order to address some of the above issues, in the first part of the dissertation we propose two empirical likelihood-based confidence intervals for the mean medical costs: One is an empirical likelihood interval (ELI) based on influence function, the other is a jackknife empirical likelihood (JEL) based interval. We prove that under very general conditions, −2log (empirical likelihood ratio) has an asymptotic standard chi squared distribution with one degree of freedom for mean medical cost. Also we show that the log-jackknife empirical likelihood ratio statistics follow standard χ2 distribution with one degree of freedom for mean medical cost. In the second part of the dissertation, we propose an influence function-based empirical likelihood method to construct a confidence region for the vector of regression parameters in mean cost regression models with censored data. The proposed confidence region can be used to obtain a confidence interval for the expected total cost of a patient with given covariates. The new method has sound asymptotic property (Wilks Theorem). In the third part of the dissertation we propose empirical likelihood method based on influence function to construct confidence intervals for quantile medical costs with censored data. We prove that under very general conditions, −2log (empirical likelihood ratio) has an asymptotic standard chi squared distribution with one degree of freedom for quantile medical cost. Simulation studies are conducted to compare coverage probabilities and interval lengths of the proposed confidence intervals with the existing confidence intervals. The proposed methods are observed to have better finite sample performances than existing methods. The new methods are also illustrated through a real example.

With new tests being developed and marketed, the comparison of the diagnostic accuracy of two continuous-scale diagnostic tests are of great importance. Comparing the partial areas under the receiver operating characteristic curves (pAUC) is an effective method to evaluate the accuracy of two diagnostic tests. In this thesis, we study the semi-parametric inference for the difference between two pAUCs. A normal approximation for the distribution of the difference between two pAUCs has been derived. The empirical likelihood ratio for the difference between two pAUCs is defined and its asymptotic distribution is shown to be a scaled chi-quare distribution. Bootstrap and empirical likelihood based inferential methods for the difference are proposed. We construct five confidence intervals for the difference between two pAUCs. Simulation studies are conducted to compare the finite sample performance of these intervals. We also use a real example as an application of our recommended intervals.

The correlation coefficient (CC) is a standard measure of the linear association between two random variables. The CC plays a significant role in many quantitative researches. In a bivariate normal distribution, there are many types of interval estimation for CC, such as z-transformation and maximum likelihood estimation based methods. However, when the underlying bivariate distribution is unknown, the construction of confidence intervals for the CC is still not well-developed. In this thesis, we discuss various interval estimation methods for the CC. We propose a generalized confidence interval and three empirical likelihood-based non-parametric intervals for the CC. We also conduct extensive simulation studies to compare the new intervals with existing intervals in terms of coverage probability and interval length. Finally, two real examples are used to demonstrate the application of the proposed methods.

The receiver operating characteristic (ROC) curve methodology is the statistical methodology for assessment of the accuracy of diagnostics tests or bio-markers. Currently most widely used statistical methods for the inferences of ROC curves are complete-data based parametric, semi-parametric or nonparametric methods. However, these methods cannot be used in diagnostic applications with missing data. In practical situations, missing diagnostic data occur more commonly due to various reasons such as medical tests being too expensive, too time consuming or too invasive. This dissertation aims to develop new nonparametric statistical methods for evaluating the accuracy of diagnostic tests or biomarkers in the presence of missing data. Specifically, novel nonparametric statistical methods will be developed with different types of missing data for (i) the inference of the area under the ROC curve (AUC, which is a summary index for the diagnostic accuracy of the test) and (ii) the joint inference of the sensitivity and the specificity of a continuous-scale diagnostic test. In this dissertation, we will provide a general framework that combines the empirical likelihood and general estimation equations with nuisance parameters for the joint inferences of sensitivity and specificity with missing diagnostic data. The proposed methods will have sound theoretical properties. The theoretical development is challenging because the proposed profile log-empirical likelihood ratio statistics are not the standard sum of independent random variables. The new methods have the power of likelihood based approaches and jackknife method in ROC studies. Therefore, they are expected to be more robust, more accurate and less computationally intensive than existing methods in the evaluation of competing diagnostic tests.

The estimation of low income proportions plays an important role in comparisons of poverty in different countries. In most countries, the stability of the society and the development of economics depend on the estimation of low income proportions. An accurate estimation of a low income proportion has a crucial role for the development of the natural economy and the improvement of people's living standards. In this thesis, the Jackknife empirical likelihood method is employed to construct confidence intervals for a low income proportion when the observed data had missing values. Comprehensive simulation studies are conducted to compare the relative performances of two Jackknife empirical likelihood based confidence intervals for low income proportions in terms of coverage probability. A real data example is used to illustrate the application of the proposed methods.

In this thesis, we apply smoothed empirical likelihood method to investigate confidence intervals for the receiver operating characteristic (ROC) curve with right censoring. As a particular application of comparison of distributions from two populations, the ROC curve is constructed by the combination of cumulative distribution function and quantile function. Under mild conditions, the smoothed empirical likelihood ratio converges to chi-square distribution, which is the well-known Wilks's theorem. Furthermore, the performances of the empirical likelihood method are also illustrated by simulation studies in terms of coverage probability and average length of confidence intervals. Finally, a primary biliary cirrhosis data is used to illustrate the proposed empirical likelihood procedure.

Compare the accuracy of two continuous-scale tests is increasing important when a new test is developed. The traditional approach that compares the entire areas under two Receiver Operating Characteristic (ROC) curves is not sensitive when two ROC curves cross each other. A better approach to compare the accuracy of two diagnostic tests is to compare the areas under two ROC curves (AUCs) in the interested specificity interval. In this thesis, we have proposed bootstrap and empirical likelihood (EL) approach for inference of the difference between two partial AUCs. The empirical likelihood ratio for the difference between two partial AUCs is defined and its limiting distribution is shown to be a scaled chi-square distribution. The EL based confidence intervals for the difference between two partial AUCs are obtained. Additionally we have conducted simulation studies to compare four proposed EL and bootstrap based intervals.

This paper uses Maximum likelihood estimation method to estimate the common correlation coefficients for multivariate datasets. We discuss a graphical tool, Q-Q plot, to test equality of the common intraclass correlation coefficients. Kolmogorov-Smirnov test and Cramér-von Mises test are used to check if the intraclass correlation coefficients are the same among populations. Bootstrap and empirical likelihood methods are applied to construct the confidence interval of the common intraclass correlation coefficients.

In this thesis, we study two methods for inference of parameters in the accelerated failure time model with right censoring data. One is the Wald-type method, which involves parameter estimation. The other one is empirical likelihood method, which is based on the asymptotic distribution of likelihood ratio. We employ a monotone censored data version of Kendall estimating equation, and construct confidence intervals from both methods. In the simulation studies, we compare the empirical likelihood (EL) and the Wald-type procedure in terms of coverage accuracy and average length of confidence intervals. It is concluded that the empirical likelihood method has a better performance. We also compare the EL for Kendall’s rank regression estimator with the EL for other well known estimators and find advantages of the EL for Kendall estimator for small size sample. Finally, a real clinical trial data is used for the purpose of illustration.