Date: Nov 22, 2012 12:03 AM
Author: Ray Koopman
Subject: Re: The problematic regression is actually ln(c) on ( ln(u), ln(u^2)<br> ), not c on (u, u^2)
On Nov 21, 8:31 pm, djh <halitsk...@att.net> wrote:
> With respect to the question I asked in the top-post, I need to note
> that the regression is actually
> ln(c) on ( ln(u), ln(u^2) )
> so one IV is twice the other, not the square of the other.
The problem is that the predictor set is singular:
one variable is a linear function of the other.
You need to square ln(u), not u.