Let user specify an informative prior

[gwern]

I was using BFA for a quick analysis of a Candy Japan A/B test, replacing prop.test with a Bayesian test, and then demonstrating the difference that an informative prior on small effect sizes (in this case dbeta(900, 1407)) makes. The model.code functionality is very nice and convenient, but it still requires something like 18 lines to extract and copy-paste and isn't very friendly to the reader. It would be nice if I could do something like

fit1 <- bayes.prop.test(c(175, 168), c(442, 439))
# with informative prior:
fit2 <- bayes.prop.test(c(175, 168), c(442, 439), prior="dbeta(900,1407)")

And a prior parameter gets substituted in the JAGS model code. (I peeked at the implementations but it looks like it bottoms out at something called jags_binom_test which isn't available.)

By the way, there seems to be a typo in bayes.prop.test where it prints "Bayesian First Aid poportion test" instead of "Bayesian First Aid proportion test".

[rasmusab]

I currently have no plans to include prior specification. But I encourage you to wrap your modified model code into a function for reuse! :)