提取Rme中lme4模型随机效应的后验估计和可信区间 [英] Extract posterior estimate and credible intervals for random effect for lme4 model in R

问题描述

我需要从模型中提取随机效应的后验估计和间隔.

I need to extract the posterior estimates and intervals for a random effect from my model.

出于说明目的，与我正在使用的数据集类似的数据集将是基数R中的 ChickWeight 数据集.

For illustrative purposes, a similar dataset to the one I am using would be the ChickWeight dataset in base R.

为固定效果提取后验估计和间隔的方法如下:

The way I extract the posterior estimates and intervals for my fixed effects is like so:

#load package
library(lme4)

#model
m.surv<-lmer(weight ~ Time + Diet + (1|Chick), data=ChickWeight)

#load packages
library(MCMCglmm)
library(arm)

#set up for fixed effects
sm.surv<-sim(m.surv)
smfixef.surv=sm.surv@fixef
smfixef.surv=as.mcmc(smfixef.surv)

#which gives
> posterior.mode(smfixef.surv)
(Intercept)        Time       Diet2  ... 
  8.5963329   8.7034260   5.1220436  ...
> HPDinterval(smfixef.surv)
                   lower      upper
(Intercept) -0.90309142 21.3617805
Time         8.42279728  9.0306337
Diet2       -6.84371527 35.1745980
...
attr(,"Probability")
[1] 0.95
>

当我为随机效果( Chick )尝试此操作时，在第二行代码中出现以下错误:

When I try this for the random effect (Chick), I get the following error at the second line of code:

smranef.surv=sm.surv@ranef
smranef.surv=as.mcmc(smranef.surv)

mcmc.list(x)中的错误:参数必须是mcmc对象

Error in mcmc.list(x) : Arguments must be mcmc objects

关于如何修改代码以提取这些值以产生随机效果的任何建议?

其他用户的注意事项:如果该模型将是MCMCglmm模型，则可以像这样提取用于随机效果的MCMC输出的后验模式值:

posterior.mode(sm.surv$VCV[,1])
HPDinterval(sm.surv$VCV[,1])

推荐答案

要提取随机效应的估算值和95％CI，请使用以下代码:

To extract the estimate and 95% CI for your random effects, you use the following code:

sm.surv <-sim(m.surv)

#between Chick variance
bChick <-sm@ranef$Chick[,,1]
bvar<-as.vector(apply(bChick, 1, var)) #between ind variance posterior distribution
bvar<-as.mcmc(bvar)
posterior.mode(bvar) #mode of the distribution
HPDinterval(bvar)

这将为您提供:

>     posterior.mode(bvar)
     var1 
     501.24353 
>     HPDinterval(bvar)
      lower   upper
var1 412.36042 630.201
attr(,"Probability")
[1] 0.95

这意味着估计值为501，较低的95％间隔为412，较高的95％间隔为630.

This means that the estimate is 501 and the lower 95% interval was 412 and the upper 95% interval was 630.

这篇关于提取Rme中lme4模型随机效应的后验估计和可信区间的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！