Just a guess, but what people don’t regularly use – they lose.

But for their audience they should just demonstrate Bayesian analyses through naive ABC (two stage sampling), pointing out the re-weighting of the prior with percent of times observations were (closely) matched, switch to re-weighting by likelihood when available and then nth roots of likelihood as a simple sequential importance sampler.

Its just a lot of grunt work to do it – adjusting it as one discovers how it can be miss-understood but the real barrier I think is giving up linear algebra and calculus as everyone should know and use that stuff all the time.

Keith

]]>