Random Effects Modelling and Inference for Skewed Data of Complex Correlation Structures

在竞争性群落中，同一地点不同种类植物或动物的观察频率通常呈负相关，而由于共享环境和其他潜在的群落特征，居住在同一群落中的人们的生活经历往往呈正相关。对相关时空变化的调查是许多环境和健康研究的核心部分。我的研究兴趣在于发展新的随机效应模型，其中随机效应可以用来解释社区效应和邻近社区之间的关联。我们的模型尤其可以通过合理的推理来解释。该方法最大的优点是对复杂依赖结构的海量数据的计算效率高。我们正在进行的研究项目包括：毒理学研究中乙二醇对出生体重和畸形的影响、芝加哥交通事故的时空分布、安大略省百湖千鱼的汞检测、美国的空气污染和死亡率、中国的地震伤亡预测和加拿大的森林火灾。我们的研究结果将有助于决策者在地理上目标资源分配和制定区域规划，以尝试修复问题区域。

Observed frequencies for different species of plants or animals at the same location are usually negatively correlated in a competitive community, whereas the life experiences of people residing in the same community tend to be positively correlated because of shared environment and other underlying community characteristics. The investigation of relevant changes over time and space is a central part of many environmental and health studies. My research interests lie in the development of new random effects models where random effects can be used to account for community effects and the association among neighbouring communities. Our models are especially interpretable with well justified inferences. The most distinctive advantage of our method is its computational efficiency for massive data of complex dependence structure. Ongoing applications of our research projects include effects of ethylene glycol on birth weight and malformation in toxicological studies, spatial and temporal distribution of traffic accidents in Chicago, mercury detection in thousand fish from hundred lakes in Ontario, air pollution and mortality across USA, earthquake casualty prediction in China and forest fires across Canada. Our findings will help policy makers geographically target resource allocation and make regional planning in an attempt to remediate problem areas.