PDA

View Full Version : probability channels



kajamix
07-11-2010, 11:38 AM
We have a weather station somewhere in the country, with the aid of satelitte information and it gives us the probability of rain every day.
Suppose it is an average value p, for the season.

At some other place there is an Indian chief who used to do the job for us for hundreds of years until such time as he was overtaken by modern technology and suppose he assigns the probability q to rain.

We know that the modern weather station is more reliable so we are likely to discard the value q and we say that p is correct.
But is that altogether true ?
Could it be that the correct probability in such situations is some function f(p,q) ? And if so how do we express it ?

SadieKhan
07-11-2010, 02:14 PM
I do not know if I am correct or not, but I would go with the number of times weather station was correct and the number of times Indian chief was correct.
Lets say chief was correct 6 out of 10 times, and weather station was right 7/10 times
maybe then we can define a function.
I am still learning probability and I would like to know what anyone else thinks.
Sadie

kajamix
02-07-2011, 08:54 PM
There is more to it than meets the eye in this one.
See here:

http://www.mathpages.com/home/kmath267.htm

The author of mathpages investigates the issue to some depth.
But what happens if we have correlations also present ?
Any more references ?

kajamix
02-07-2011, 09:00 PM
mathpages suggests the answer:

f(p,q) = p*q / (p*q+(1-p)*(1-q))

on the assumption that the separate predictions are uncorrelated.
But some correlation is likely to exist in such situations, which in the case of modern weathermen v. Indian chief might be the identicality of certain methods used by both.