This dissertation, "Semiparametric Methods in Generalized Linear Models for Estimating Population Size and Fatality Rate" by Danping, Liu, 劉丹平, was obtained from The University of Hong Kong (Pokfulam, Hong Kong) and is being sold pursuant to Creative Commons: Attribution 3.0 Hong Kong License. The content of this dissertation has not been altered in any way. We have altered the formatting in order to facilitate the ease of printing and reading of the dissertation. All rights not granted by the above license are retained by the author. Abstract: Abstract of thesis entitled SEMIPARAMETRIC METHODS IN GENERALIZED LINEAR MODELS FOR ESTIMATING POPULATION SIZE AND FATALITY RATE Submitted by LIU Danping for the degree of Master of Philosophy at The University of Hong Kong in August 2005 Thisthesisinvestigatestheextensionsandapplicationsofgeneralizedlinear models by incorporating nonparametric and semiparametric frameworks. The nonparametrictoolsallowtheestimationofanonlinearcurve, whilesemipara- metric methods are used to separate the e(R)ect of covariates from unexplained trend e(R)ects. Kernel smoothing methods were used to extend the Poisson log linearapproachtotheestimationofthesizeofpopulationusingmultipleliststo an open population when the multiple lists were recorded at each time point. The approach is based on a two-step procedure. In the rst step log-linear models were used to estimate the mean number of unlisted addicts at each time point. The estimated total number of addicts, which is the sum of the listed and unlisted addicts, is supposed to be a smooth function of time and a kernel smoothing method was then applied to produce smooth estimates. Thee(R)ect of exogenous factors was considered using a semiparametric model. A prole kernel estimating equation approach was adopted to obtain the explicit expressionoftheinterestedparametersaswellastheirstandarderrors. Asim- ulation study revealed the proposed estimators and their estimated variances give reliable estimates and the resulting approximate condence intervals are close to their nominal coverage probabilities. A model for an age and time dependent fatality rate of a disease was proposed. The age e(R)ect recognizes that individuals of di(R)erent ages may have di(R)erent fatality rates, but these were assumed to vary smoothly with age. The time dependence allows for timedependente(R)ectssuchaslearninge(R)ectsinthetreatmentofdiseasesand increased fatality rates when the hospital system is chaotic, especially in the early part of an emerging epidemic. DOI: 10.5353/th_b3616459 Subjects: Population - Statistical methodsMortality - Statistical methodsParameter estimation