On 2004-07-10, Bryan J. Maloney wrote:
Martin Brown abagooba zoink larblortch
But I can still compute a function describing the probability
distribution of believing in the probability of a second example being
found P(P(life)) given all the available data. That is in essence what
Bayesian statistics does. And it is not hamstrung by the N=1 situation.

And how, given N=1 is this substantially better than a wild-ass guess?

Because it tells you exactly how limited this wild-ass guess is.

It has certain nasty properties with such very limited data. Most
notably that it cannot be normalised so the variance in any individual
estimate P(life) is infinite. However, it represents accurately our
current state of knowledge.

So does "We've only had one observation, so no meaningful inference can
be drawn." I hadn't realized that Bayesians could be so cultlike.

Meaningful, but limited inference can be drawn.

--
Aaron Denney
--