b******2 发帖数: 654 | 1 两个group,用propensity score matching得到 500 病人, 1000个健康人, 应该用什
么test 比较matched group的 continuous and binary outcomes?
1:1 propensity score matching will need to use paired t-test/signed rank
test for continuous, McNemar for binary outcome.对1:2 matched , 好像不能用
。 谢谢! | y*****w 发帖数: 1350 | 2 Need to use specific analysis methods for paired data as you pointed out,
because the matched pairs are not independent anymore. Would be
complicated if you have to run mixed or survival models. Check out
Peter Austin's papers.
Would suggest using the weighting method (IPTW) instead of whatever matching. | k****n 发帖数: 165 | 3 如果1000个健康人都是unique的(没有用一个健康人match2个或以上病人的情况),可
以take average of the 2 matched control observations within each pair.
Besides,why use 1:2? Ppl use 1:K when they have a rich control set but
usually K is much larger than 2. Do pay attention the balance of the
propensity score in each matched pair for observations from treatment and
control group. This is the advantage of using matching versus inverse
weighting as it is obvious for you to check whether you have good matched
pairs. For binary treatment, matching, subclassfication, inverse weighting
should have about same performance. It's another story if the treatment is
continuous. |
|