Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Picking a model with confidence

0 views
Skip to first unread message

AngleWyrm

unread,
May 11, 2008, 2:46:21 AM5/11/08
to
Let's say we have a random variable X, and two functions f(X) and g(X). Both
functions produce a success/failure result, but they do not have exactly the
same probability of success.

We are given an unknown function h(X), which could be either f(X) or g(X).

After many trials, the average number of successes from h(X) will converge
on one or the other. It will become apparent that either f(X) or g(X) is
being used.

How many trials are needed to differentiate between them with say 99%
confidence?


0 new messages