Your browser does not support JavaScript!
2017-10-20(五),主講人:顏佐榕 教授 (中央研究院 統計學研究所)


統 計 學 研 究 所

專 題 演 講


講 題: Covariance Matrix Estimation in Random Effect Models Using Metropolis Stochastic Gradient Algorithms with Data-driven Proposals
演講者: 顏佐榕教授 (中央研究院 統計學研究所)
時 間: 106年10月20日(星期五)10:40 - 12:00noon  (10:20 - 10:40am 茶會於統計所821室舉行)
地 點: 綜合三館837室
摘 要:

In this paper we propose a method for estimating the covariance matrix of random effects in mixed effect models. We adopt a stochastic approach to approximating the gradient of the log marginal likelihood function. This approach uses a Metropolis-Hastings algorithm to sample random effects from distributions conditional on observed information. In particular, to make the algorithm efficient, this approach uses a data-driven proposal for sampling random effects from target distributions. As a result of that, this approach can achieve high accuracy in approximating the log marginal likelihood. We apply the approach to estimate the covariance matrix of the random effects in various models, including bivariate joint models and the Gaussian graphical model.


敬請公佈                                 歡迎參加

  • 友善列印
  • 分享
Voice Play