Empirical Likelihood Methods In Biomedicine And Health PDF Download

Missing covariate data occurs often in regression analysis, which frequently arises in the health and social sciences as well as in survey sampling. This dissertation contains three topics in nonignorable covariate-missing data problems, in which we study methods for the analysis of a nonignorable covariate-missing data problem in an assumed conditional mean function when some covariates are completely observed but other covariates are missing for some subjects. First, by exploitation of a probability model of missingness and a working conditional score model from a semiparametric perspective, we propose a unified approach to constructing a system of unbiased estimating equations, where there are more equations than unknown parameters of interest. These unbiased estimating equations naturally incorporate the incomplete data into the data analysis, making it possible to seek efficient estimation of the parameter of interest even when the working regression function is not specified to be the optimal regression function. Based on the proposed estimating equations, we introduce three maximum empirical likelihood estimators of the underlying regression parameters and compare their efficiencies with other existing competitors. By utilizing the proposed empirical likelihood method on a data set from the US National Health and Nutrition Examination Survey (NHANES), we study the effect of daily alcohol consumption on hypertension. Second, we explore unconstrained and constrained empirical likelihood ratio statistics to construct empirical likelihood confidence regions for the underlying regression parameters without and with constraints. We establish the asymptotic distributions of the proposed empirical likelihood ratio statistics. The proposed empirical likelihood methods have a better finite-sample performance than other existing competitors in terms of coverage probability and interval length. An analysis on the data set from the US NHANES demonstrates that increased alcohol consumption per day is significantly associated with increased systolic blood pressure. In addition, higher body mass index and older age have a significantly higher risk of hypertension. Third, we propose a pseudo empirical likelihood ratio statistic, yet it is demonstrated following an asymptotically chi-squared distribution. Our proposed method allows for confidence interval construction without variance estimation and thus is more computationally feasible. Simulation results suggest that the proposed empirical likelihood confidence interval has a better finite-sample performance than the corresponding Wald-based competitor in terms of coverage probability and interval length. Moreover, the proposed empirical likelihood ratio test is always superior to the Wald method in terms of their power performances in our simulation studies.