home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!noc.near.net!hri.com!spool.mu.edu!uwm.edu!linac!att!cbnewse!cbnewsd!lew
- From: lew@cbnewsd.cb.att.com (lewis.h.mammel..jr)
- Newsgroups: sci.math
- Subject: Re: Lagrange successor rule
- Summary: bayesian probability
- Message-ID: <1992Dec23.031720.7464@cbnewsd.cb.att.com>
- Date: 23 Dec 92 03:17:20 GMT
- References: <3207@devnull.mpd.tandem.com>
- Organization: AT&T
- Lines: 41
-
- In article <3207@devnull.mpd.tandem.com>, garyb@mpd.tandem.com (Gary Bjerke) writes:
- >
- > I was thumbing through an old statistics textbook when I came across the
- > Lagrange successor rule. The example given was that of a collection of coins
- > for which the probability of flipping a head is uniformly distributed over
- > the set of values {1/N, 2/N, ..., N/N} for N some arbitrary integer. The rule
- > states that the probability of getting a head on the (n+1)th flip given that
- > the first n flips were heads, is n/(n+1).
- >
- > Note that N is arbitrary and not a parameter is the final result. The textbook
- > said Lagrange had used this to prove that the sun was extremely likely to rise
- > tomorrow, given that it had risen every day for the past 5000 years (the
- > probability is [5000*365]/[5000*365+1], to be precise).
- >
- > I followed the proof, but I have absolutely no intuition for this result at
- > all. I even fail to see how it applies to the rising of the sun (in what sense
- > does the unconditional probability of its rising meet the uniform-distribution
- > requirements?) Can somebody help me get a gut feel for what this result means?
-
- It's supposed to model your a priori expectations, which are taken
- to be extremely open-minded. Part of the idea is that after evidence
- accumulates the conditional expectation will depend weakly on the
- exact a priori distribution, as long as its very broad.
-
- Conversely, if I insist that I KNOW ( for example ) that an
- atom has a probability per unit time of decaying of 100/sec, but
- I've been watching it for a week and it's still there, I'm within
- the bounds of rationality if I maintain my claim. Realistically
- though, I would probably start to develop doubts, since my certainty
- would force me to accept that an event with a probability of
- exp(-60480000) ( and getting exp(-100) smaller every second ) has
- occurred.
-
- This means I should have modeled my a priori expectation of the
- intrinsic decay rate with a strong peak at 100/sec, and some broad
- background. As time goes on without a decay, the longer decay times
- will quickly gain strength in my conditioned expectations, even if
- I was 99.9999999999999999999999999999999999% sure of that 100/sec
- to start with.
-
- Lew Mammel, Jr.
-