confidence interval,统计讨论

请问大家如何解释两个/3 个 confidence interval 呢，是要考虑bonferroni correction，还是怎么样呢，
尤其是一个metrics 正向显著，另一个负向显著，要和product产品联系在一起分析原因，有没有思路呢

The more things you test, the more likely you are to see significant differences just by chance (false positive).
-> We should adjust the significance level, so it accounts for how many metrics or how many different tests you're doing

I think Bonferroni correction is one way, but it's too strict.
Another way is to check family-wise error rate FWER, control false discovery rate FDR, but only use this when you have a lot of metrics
Also check Holm–Bonferroni method, 但我对这个method具体怎么操作不太熟

波屯穗香 发表于 2019-12-15 01:02
The more things you test, the more likely you are to see significant differences just by chance (fal ...

but how to interpret when these metrics have different direction, one is positive another is negative, are they both significant?

product sense必备，找人合买

多少钱啊。

akiii 发表于 2019/12/17 12:43:21
多少钱啊。

我看好像是299一套，对吗

一起买啊，原版的带数据吧？

what is this mean?

crity 发表于 2019-12-19 01:05
一起买啊，原版的带数据吧？

对的，带数据和PDF

我们两个一起买啊， 回头有加入的再随便分分？ 先一个人150