|
BGonline.org Forums
math/stat question (OT)
Posted By: Chuck Bower
Date: Saturday, 23 July 2011, at 5:42 p.m.
(Not likely I'm using the 'preferred' symbolism/nomenclature, but I think you'll understand.)
Suppose you take some data and suspect a relationship between x (some independent variable) and P(x) (the probability that something is true, given x). You have several measurements of xi and p(xi) [here I use a small 'p' to distinguish it as a measurement as opposed to the supposed TRUE relationship, which is unknown]. You have several different values {xi for i = 1,2,3...} and the corresponding measurements pi(xi). But for each value of i you have a different number of measurements.
So you find a mean p-bar(xi) for all i and plot them up. Now you want to fit a curve to them ("curve" could be a straight line). How do you weight the different p-bar(xi)?
|
BGonline.org Forums is maintained by Stick with WebBBS 5.12.