USING THE EXPONENTIAL TILTING MODEL AND THE MONOTONIC DENSITY RATIO MODELTO FIND THE EMPIRICAL DISTRIBUTION FOR PARAMETERS

Abstract

With the development of technology, the clinical study related to disease detection has become a popular topic for biostatisticians. For those studies, the combination of multiple diagnostic tests seems to be the most general tool to achieve optimal sensitivity and specificity. In this thesis, we apply the exponential tilting model proposed by Qin and Zhang (2010) and the monotonic density ratio model proposed by Chen et al. (2015) to estimate the asymptotic distribution for β and AUC using three bootstrapping methods and give an evaluation and comparison towards their performances. We give a good estimation for the distribution of β no matter whether the robustness is taken into consideration. And we also have a good thinking for estimating the distribution of AUC.