主题：学术讲座预告(更新)

应杨院长邀请，加拿大Waterloo大学统计与精算系（Department of Statistics and Actuarial Science）Peter SONG博士，拟于2006年4月7日来我院讲学。讲学题目为“Statistical Computing”（统计计算），欢迎大家到时踊跃参加。

时间：2006年4月7日上午10点（暂定）

地点：理科楼207

报告内容：

Vector Generalized Linear Models: A Gaussian Copula Approach

Abstract

I will present a multivariate extension to the classic generalized linear models using the approach of the Gaussian copula. This extension furnishes a unified framework of analyzing multi-dimensional nonnormal data in the context of regression analysis. The proposed models can accommodate discrete, continuous and mixed vector outcomes arising in many practical studies, including longitudinal, clustered or spatial data. With the availability of joint probability distributions, we develop a simultaneous maximum likelihood inference in the class of models, where a Gauss-Newton type algorithm is implemented to obtain numerical results for the parameter estimation. I will illustrate vector logistic (or probit) models for multivariate binary outcomes, vector log-linear models for multivariate count outcomes, and a vector GLM for normal and binary mixed outcomes. Comparisons to the GEEs (Generalized Estimating Equations) approach will be discussed in detail through both simulation and data analysis examples.

不好意思，没有翻译出来（水平有限）。也许原汁原味更好吧^_^。