Statistical Research on Covariate Matching, Monotone Functional Data and Binary Spatio-temporal Data Modeling

This dissertation consists of three studies in different fields. (1) The first study aims to evaluate the effect of wind turbine upgrades by devising a covariate matching method. The proposed method performs straightforward comparison between treated and non-treated outputs by re-organizing data records as if they were designed by randomized experiment. Also, it considers multi-dependencies of dynamic atmosphere conditions by taking into account the priority order and interaction effect of factors. (2) The second study proposes the functional data model to estimate a collection of monotone curves observed on an irregular and sparse grid. By integrating functional principal component analysis, not only does the model describe the variation of curves by few important functions but also it jointly estimates numerous monotone curves having a mean trend as well as individual-specific features. Simulation study validates its superiority to other classical approaches. (3) The last study investigates spatio-temporal binary data with a goal of describing infectious disease spreading pattern. An autologistic regressive model is proposed to illustrate spatial dependence and predict the progression over space and time. The accuracy of model estimation is verified by simulation study. Additionally, a hidden Markov network model is established from a slightly different standpoint on given data