...
【24h】

Letter to the Editor

机译:给编辑的一封信

获取原文
获取原文并翻译 | 示例
           

摘要

Due, Jalaludin & Morgan (2009) (DJM) conclude that 'In all cases, the BMA [Bayesian model averaging] results should be used rather than those of the single model using either the Bayesian method or classical GLM or GAM methods.' I accept that it is better to average over a set of plausible models. Relying on a single model leads to credible intervals that are too narrow because model uncertainty is ignored (Gelman & Rubin 1995). But DJM seem to imply that because estimates differ between their hierarchical Bayesian model and BMA, the latter estimates are to be preferred (Due ef ai. 2009, p. 298, 302) and I am not convinced this is so. First their hierarchical Bayesian model can also be interpreted as a model for uncertainty about the relationship between a Poisson rate parameter and a set of covariates (Albert 1988). Using the notation x ~ D[a,b] to represent a random variable x distributed D with mean a and variance b, the first two levels of their hierarchical model can be re-written as:
机译:因此,Jalaludin&Morgan(2009)(DJM)得出结论:“在所有情况下,都应使用BMA(贝叶斯模型平均)结果,而不是使用贝叶斯方法或经典GLM或GAM方法的单个模型的结果。”我接受最好对一组合理的模型进行平均。依靠单一模型会导致可信区间过窄,因为模型不确定性被忽略了(Gelman&Rubin 1995)。但DJM似乎暗示,由于估计的层次贝叶斯模型与BMA之间存在差异,因此最好使用后者的估计(Due ef ai。2009,p。298,302),但我不相信这是事实。首先,它们的分层贝叶斯模型也可以解释为泊松速率参数与一组协变量之间的不确定性模型(Albert 1988)。使用符号x〜D [a,b]表示具有均值a和方差b的随机变量x分布D,可以将其层次模型的前两个级别重写为:

著录项

  • 来源
    《Australian & New Zealand journal of statistics》 |2011年第2期|p.257-259|共3页
  • 作者

    Jim Young;

  • 作者单位

    Department of Public Health and General Practice University of Otago, Christchurch, New Zealand;

  • 收录信息
  • 原文格式 PDF
  • 正文语种 eng
  • 中图分类
  • 关键词

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