在完全匹配后使用 Survey::svydesign 和 tbl_svysummary 来考虑匹配权重是否有意义？
我使用 R 中的 MatchIt 包进行完全匹配，它为我提供了匹配数据集中每个观察值的权重。然后我想列出治疗组和对照组的结果汇总表。
我无法使用 Sjoberg 的 gtsummary 包中的正常 tbl_summary，因为它不考虑匹配权重。所以，我想我可能会使用 tbl_svysummary （调查包），因为它可以让我考虑匹配权重。
因此，我的代码如下所示：
Extract<-matched_data %>% select(outcome, Treatment, Weights)
Survey::svydesign(~1, data = as.data.frame(Extract), Weights = ~weights) %> ;% tbl_svysummary(by=Treatment,digits = list(all_categorical() ~ c(0, 2))) %>% add_p()
我的问题：这有意义吗？或者如何在完全匹配后呈现一个汇总表，以便我能够说明匹配权重？
谢谢您的任何建议！
I do full matching with the MatchIt package in R, which gives me the weights for each observation in the matched dataset. I then want to present a summary table of the outcomes for the treatment and the control group.
I cannot use the normal tbl_summary from the gtsummary package by Sjoberg as it does not account for matching weights. So, I thought I might use tbl_svysummary (survey package) because it would allow me to account for matching weights.
Thus, my code would look like:
Extract<-matched_data %>% select(outcome, Treatment, weights)
survey::svydesign(~1, data = as.data.frame(Extract), weights = ~weights) %>%
tbl_svysummary(by=Treatment,digits = list(all_categorical() ~ c(0, 2))) %>% add_p()
My question: Does that make sense? Or how can I present a summary table after full matching that allows me to account for the matching weights?
Thank you for any advice!
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。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
您可以这样做，但是，我定期处理调查数据，根据我的经验，当我想要呈现汇总表时，使用 Survey::svymean() 或 Survey::svyby() 函数更容易。
例如使用 svyby：
svyby(~Treatment, ~outcome, design = YOURDESIGN, svymean)请参阅文档https://www.rdocumentation.org/packages/survey/versions/4.1-1/topics/svyby
这本书“复杂” Lumley 的“调查” 也是一个很好的资源，可以阅读和了解有关调查设计的更多信息，并以简单的方式总结结果。
You can go about it that way, however, I work with survey data regularly and in my experience its easier to use the survey::svymean() or survey::svyby() functions when I want to present summary tables.
So for example using svyby:
svyby(~ Treatment, ~ outcome, design = YOURDESIGN, svymean)Please refer to the documentation https://www.rdocumentation.org/packages/survey/versions/4.1-1/topics/svyby
The book 'complex surveys' by Lumley is also a great resource to read and learn more about survey designs and also summarising your results in easy ways.