Bob Carpenter

I do think it'd make sense to test the other internal functions, which I haven't done yet. And if we do need to test the internal algorithm details, then it'd...

Given @WardBrian's comments about typing, I'll try to get the ESS updates in first, which will in turn make it easier to write a generic testing framework, which will in...

> it's me again being a pain in the arse. I like feedback, so don't worry. > one or more of sens, spec and p (\pi in the paper) have...

Hmm. I thought I replied to this, but looks like I never hit "Comment". Argh. It was a long reply. Let's start again. You don't need to start your own...

> I wonder if it also makes sense to simulate from "the prior" by just omitting the likelihood on the tests, i.e., including the sens/spec data: Yes, and you can...

> For instance one can simulate the prior and posterior predictive of the number of positive tests and those are going to be regular histograms, right. Yes, good point. We...

*[Edit: see below; I think this is answering the wrong question]* If you fit the Poisson model to a set of data, you can compute the predictive of mean(y_rep) and...

Oh, I think what I said was addressing a different question about the posterior predictive. If you just do prior predictive checks or posterior predictive checks, each data set generated...

I agree that a U-shaped prior on prevalence doesn't make sense. But it's flat enough I doubt it has a substantial impact on the posterior.

> Not sure if I missed something, but these two are the same? I'm blaming the spell checker, which changed "exponantiate" back to "exponentiate".