• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Statistical Methods and Applications >Discussion of 'Sampling schemes for generalized linear Dirichlet process random effects models' by M. Kyung, J. Gill and G. Casella
【24h】

Discussion of 'Sampling schemes for generalized linear Dirichlet process random effects models' by M. Kyung, J. Gill and G. Casella

机译：M. Kyung，J。Gill和G.Casella对“广义线性Dirichlet过程随机效应模型的采样方案”的讨论

获取原文
获取原文并翻译 | 示例
       
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

The paper evaluates new and existing MCMC strategies for Generalized Linear Mixed Dirichlet Process Models (GLMDMs), specifically, logistic, probit and log-linear GLMDMs with normal base measure. Gibbs samplers for linear models and pro-bit regression models (Albert and Chib 1993) have, of course, been around for several years. Slice samplers or Metropolis-Hastings (M-H) strategies are typically used for analyzing logistic and log-linear models. An important contribution of this paper is that it utilizes a representation of the logistic distribution as a normal mixture of the Kolmogoriv-Smirnov distribution (Andrews and Mallows 1974; West 1987) to extend Gibbs sampling to logistic regression models. For log-linear models, the authors recommend an M-H algorithm where the proposal distribution corresponds to a linear mixed model fitted to the log-responses. Simulation studies demonstrate better mixing properties of the recommended algorithms relative to slice samplers by a comparison of autocorrelation function (ACF) plots.
机译：本文评估了适用于广义线性混合狄利克雷过程模型（GLMDM）的新的和现有的MCMC策略，特别是具有常规基本度量的对数，概率和对数线性GLMDM。线性模型和位回归模型的吉布斯采样器（Albert and Chib 1993）当然已经存在了几年。切片采样器或Metropolis-Hasting（M-H）策略通常用于分析逻辑模型和对数线性模型。本文的重要贡献在于，它利用逻辑分布的表示形式作为Kolmogoriv-Smirnov分布的正态混合（Andrews和Mallows 1974； West 1987），将吉布斯采样扩展到逻辑回归模型。对于对数线性模型，作者建议使用M-H算法，其中提案分配对应于适合对数响应的线性混合模型。仿真研究通过比较自相关函数（ACF）图，证明了相对于切片采样器，推荐算法的更好混合特性。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号