Galton's Bayesian Machine

[mhb]

http://chance.amstat.org/2011/02/galton/

[zephjc]

This is pretty brilliant. A simulator of this device would make a good tool teaching tool for helping people grok bayes' formula. Weekend project anyone?

[nazgulnarsil → zephjc]

Simulator? Build a real one!

[abeppu]

That's pretty awesome. Now if only we could build a multivariate version that let you estimate covariance matrices also ...

[sesqu]

What does Bayes Theorem look like? [--] I mean, how can we visualize the cognitive content of the theorem?

Umm, how about a Venn diagram? Seriously, it's not that complicated. In my mind's eye, p(µ|θ) := |{ x | µ(x)∧θ(x) }| ÷ |{ x | θ(x) }|.

[pygy_ → sesqu]

This works for the discrete version, but not for the continuous one (if your prior, likelihood and posterior are distributions).

[snth]

Any opinions or experience with the 'Chance' magazine? It looks interesting, and Andrew Gelman is an editor.