A Simulation study of Poisson and Distributed Lag Models Under the structure of Zero-inflated outcomes
| Autor: | Wun-Kai Jiang, 江文楷 |
|---|---|
| Rok vydání: | 2016 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 104 For the count data, a popular method in medicine, public health and epidemiology research is the Poisson Regression Model estimating the relative risk (RR); Polynomial Distributed Lag Model (PDLM) is also a popular strategy in environmental epidemiology research to examine how the delayed environmental factors influence infectious disease. This model assumes that the dependent variable (Yt) is not only effected by the current independent variable (Xt), but also the lagged predictors (Xt-1, Xt-2 ...); Recently Distributed Lag Non-linear Model (DLNM) has been widely adapted. The feature of this model is that- it not only considers the lag effect between the environmental factor and the infectious disease, but also takes into account the non-linear relationship. However the performance of the three models for the zero-inflated outcome are unknown. Therefore, the simulation study we based on the time association between temperature and dengue fever under various models such as Zero-inflated Poisson, Zero-inflated Negative Binomial and Normal Distribution to estimate power of these three models in difference parameter settings. We will also implement using the permutation method under null hypothesis to evaluate the type I error rate performance for Polynomial Distributed Lag Model and Distributed Lag Non-linear Model. Finally, we applied the three models to the longitudinal data from 1998 to 2008 in Kaohsiung, Taiwan to confirm the association between dengue fever and temperature. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|