Connect your moderator Slack workspace to receive post notifications:
Sign in with Slack

Mixture of Linear Regression

Hi!
I was doing the mock midterm exam of 2015, and I can't understand the solution of the last point, where you ask us to proof if the model is identifiable or not. Concerning this point, in general, when it's asked whether a model is identifiable or not, what should we do? What are the passages required?
Thanks for your answer!

There is already a problem in the solution of the question. More precisely, in the question definition of the identifiabilty, we have the inverse relationship. A model is identifiable iff P{theta1} = P{theta2} implies that theta 1= theta 2. The proof is correct as soon as if you have a condition A-> B it is the same non(B) -> non(A), and here non (B) is the fact that theta 1 ! = theta 2 and non A is the fact that P_theta1 = P_theta2. As you prove that non(B)-> A, it means that the model is not identifiable. Tell me if you have any mathematical problem with the notations I used. You can prove it by an easier way with the same idea as in the prove, by proving that A -> non(B).

Page 1 of 1

Add comment

Post as Anonymous Dont send out notification