Re: predicting outcomes was: Re: RMA is dying (Hal is an obstinately stupid putz.)

"Jerry B. Altzman" <jbaltz@xxxxxxxxxxx> wrote in
news:47be2744$0$25054$607ed4bc@xxxxxx:

No, you don't get it.

Estimation of probabilities (mathematical estimation as a number, that
is) is NOT a "one size fits all" thing. No, the (estimate of the)
probability is necessarily based on *a particular set of available data.*
That set may vary for different estimators and they will, entirely
correctly, produce different estimates based on it. It does not even
matter if the event has already happened and one outcome has resulted -
if an estimator does not have that fact in his estimating basis he will,
quite properly, assess the probability of the event not as 1.0.

The outlook that a probability varies as a function of the estimating
knowledge basis, and is therefore not a single number, traces back
(mainly) to Bayes.

Now, for certain problems, if the working knowledge base is ONLY that
there are two and only two mutually exclusive outcomes (let's call them A
and B) - then the probability of A is 50% (and B is also 50%). The
proof? It follows directly from symmetry.

However, as I said, it is a rare case indeed when this and ONLY this
knowledge base applies - we often have more info that bears, however
tangentially, on the problem.


Regards,

PS Aside from the fact that most people (including highly trained
people) find Bayesian analysis very counterintuitive, there is another
reason why people do not accept (or at least are very unconfortable with)
the 50/50 proposition applying when all that is known is that there are
two exhaustive and exclusive possibilities. That is because, besides the
explicit knowledge basis on which the analysis is doen, most people have
an implicit, "back pocket" knowledge base which they also secretly (or
not so secretly) use.

.

Relevant Pages

  • Re: Probabilty Question
    ... I think that you asked whether you can tell what the heads probability ... paramater is in the interval. ... "prior" distribution on the value of the paramater. ... Estimation is a huge area of statistice, ...
    (sci.math)
  • Re: Shannons information theory
    ... random variable has entropy 0. ... The probability distribution associated with a random variable is almost ... the observed data as the probabilities) and bayesian estimation. ... and then it adjusts this prior based on the observed data. ...
    (comp.theory)
  • Re: How to evaluate the propability distribution of two dices?
    ... probability of sum? ... p as the estimated probability. ... the number of sides of dices and the number of dices: ... and estimation of probabilities P. ...
    (sci.stat.math)
  • Re: A simple but confusing question
    ... This is not an estimation problem. ... white balls in the bucket is a nuisance parameter, ... out of the expression for the probability. ... > and a prior, which expresses your prior beliefs about the parameters. ...
    (sci.stat.math)