Clubsandwich :: conf_int在荟萃分析中 - SE公式不给出同样的顺式

发布于 2025-02-11 14:20:38 字数 755 浏览 2 评论 0原文

因此，我使用Metafor（RMA.MV）和Clubsandwich来计算CHE模型荟萃分析，但是当我使用clubsandwich :: conp_int功能来估计我的集群强大的SES和CIS时，我注意到当我在顺式中发现了一些差异。使用SE以通常的方式计算它们（lnor +/-（1.96*se）

代码为：

attbel_mods <- rma.mv(LOR ~0 + gender, 
                      v = att_var, 
                      random = ~1 | Study/es.id, 
                      test = "t", 
                      data = rattbel, 
                      method = "REML") 
 clubSandwich::conf_int(attbel_mods, vcov = "CR2")

以下是输出的样本（不允许发布屏幕截图）

coef.estimate    0.437   
se.              0.390114
d.f.             2.78 
Lower 95% CI     -0.861 
Upper 95% CI     1.73

使用正常公式来估计CI，我得到了LCI = -0.328，UCI = 1.201

我在某个地方出错了

！

任何人都可以告诉我，或者

原文

So I'm using metafor (rma.mv) and clubSandwich to compute a CHE model meta-analysis, but when I use the clubSandwich::conf_int function to estimate my cluster robust SEs and CIs, I noticed some discrepancies in the CIs when I use the SE to compute them in the usual way (lnOR +/- (1.96*SE)

Code is:

attbel_mods <- rma.mv(LOR ~0 + gender, 
                      v = att_var, 
                      random = ~1 | Study/es.id, 
                      test = "t", 
                      data = rattbel, 
                      method = "REML") 
 clubSandwich::conf_int(attbel_mods, vcov = "CR2")

Below is a sample of the output (not allowed to post a screenshot yet)

coef.estimate    0.437   
se.              0.390114
d.f.             2.78 
Lower 95% CI     -0.861 
Upper 95% CI     1.73

Using the normal formula to estimate a CI, I get LCI = -0.328 and UCI = 1.201

Can anyone tell me why this is, or if I've gone wrong somewhere!

Thank you!

Hannah

分享到QQ

分享到微博

如果你对这篇内容有疑问，欢迎到本站社区发帖提问参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。

发布评论

需要登录才能够评论，你可以免费注册一个本站的账号。

好菇凉咱不稀罕他 2025-02-18 14:20:39

在您的公式中1.96在从z-table @ 0.05显着级别计算时使用临界值，但是，此处的关键值是根据t-t-table 显然，自由度为2.78，

因此是：

coef.estimate <- 0.437
se <- 0.390114
df <- 2.78
t.score <- abs(qt(0.025,2.78)) #0.05/2 = 0.025 for each tail
LCL = coef.estimate - t.score * se
UCL = coef.estimate + t.score * se 

> t.score
[1] 3.329777
> LCL
[1] -0.8619925
> UCL
[1] 1.735993

in your formula 1.96 is used when critical values are calculated from z-table @ 0.05 significance level, however, here critical values are calculated from t-table as apparently, degrees of freedom is 2.78

so here it is:

coef.estimate <- 0.437
se <- 0.390114
df <- 2.78
t.score <- abs(qt(0.025,2.78)) #0.05/2 = 0.025 for each tail
LCL = coef.estimate - t.score * se
UCL = coef.estimate + t.score * se 

> t.score
[1] 3.329777
> LCL
[1] -0.8619925
> UCL
[1] 1.735993

回复收藏 0 原文

~没有更多了~