Gelman et al. "Bayesian Data Analysis 3rd ed." (いわゆる BDA3) p.96 つらつらとながめてるいて…たとえば 8 school とか,いわゆる多重比較は階層モデルがいい,自動的に学校間の推定値を partially pool するから(学校間の差が小さくなる)…とのこと.「まあ,そうだね」というかんじですが…
16:29:04
RT
@hwassner
: Hi
@StatModeling
I like to educate people about the multiple comparison problem when practicing AB tests (on the web). Especially I'd like to encourage the use of hierarchical bayesian testing (aka "partial pooling"). Is there any way to collaborate ?
19:48:42
RT
@UChicagoCAPP
: 'Forty years ago, there were no models in Tukey's book "Exploratory Data Analysis", since he had no understanding of how exploring data would even be related to models' argues Prof. Andrew Gelman (
@StatModeling
) 'We've come a long way since then.'
pic.twitter.com/p7bn8ewzdV
19:52:01