I don't do this stuff, so this may be a stupid question. But it looks like you have p_A and p_B being independent in the posterior distribution. How does that w

I don't do this stuff, so this may be a stupid question. But it looks like you have p_A and p_B being independent in the posterior distribution. How does that work? I would assume that they're pretty correlated IRL --- we're not changing that much in A/B testing. This must be coming from the likelihood function, so you may want to be more specific about the exact experiment you have in mind. I'm not sure off the top of my head whether imposing independence is going to overstate or understate the precision of the estimate...