Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Joint of normal and logistic

  1. Oct 8, 2011 #1

    I would be really happy to recieve some help. I have tried using Jacobians and so on, but
    I am stuck.

    I'll start with the univariate case. Let X ~ N(μ,σ) and Y = exp(X)/(1+exp(X)). What is the joint f(x,y)? According to intuition fy|x = 1, but since we are dealing with continous distributions I am not sure.

    The bivariate case is an extension. Let X1, X2 be bivariate normal and
    Y = exp(γ1*X1 + γ2*X2)/(1 + exp(γ1*X1 + γ2*X2) ), where the gammas are just
    constants. Now I am looking for f(x1,x2,y). Any suggestions of how to proceed at least?

  2. jcsd
  3. Oct 9, 2011 #2

    Just to clarify the first case. I actually want to find the joint f(x,y) and then E(XY). Since Y=g(X), then f(x, g(x)) and E(XY)= E(Xg(X)). We know that Pr(Y=exp(x)/(1 + exp(x))) = 1, but I am not sure if then can write fy|x = 1 (all probability mass concentrated in one point). If that is the case, then f(x,y) = f(x) and E(XY) = ∫∫xyf(x)dxdy = ∫xexp(x)/(1+exp(x))f(x)dx which can easily be calculated numerically. Am I on the right track?

    The bivariate case is more complicated but maybe that can be solved as well. :)

Share this great discussion with others via Reddit, Google+, Twitter, or Facebook