Hi, I am working on an randomized control trial model to determine the estimates for trial/control and a distribution (trial/control email sent on same date). I would expect the distribution beta estimates to be 0 instead of ~0.1, as the testing data controls for the distribution effect. It seems t…

I don’t think this is related to trial_control at all, but to the normal prior on the intercept. The problem goes away if I replace that with a b0 = pm.Flat('b0').

Understanding params in partially pooled model

Questions

chad39 July 29, 2019, 5:13pm 3

Thank you! That worked perfectly.

Topic		Replies	Views
Models with different pooling give very different results Questions	2	553	December 26, 2017
Simple Dirichlet model with partial pooling Questions	1	714	February 22, 2019
Prior distribution for "certainty" parameter of Beta distribution Questions	5	2058	April 10, 2018
Feedback on RCT model Questions	0	415	July 26, 2019
Parametrize mixed beta regression in non centered way Questions	6	1194	April 17, 2018

Understanding params in partially pooled model

Related topics