Page 1 of 1
log of the intercept in least squares
Posted: Fri Jul 30, 2010 8:44 am
by trout
I was estimating a power regression y=a*x^(-b), where a and b are my coefficients. The natural log of the function ln(y)=ln(a)-b*ln(x) fitted the data quite well. But I am not sure if e-views estimated this relationship with nls or ols. More generally, what does eviews actually do when you take the log of the intercept?
Re: log of the intercept in least squares
Posted: Fri Jul 30, 2010 8:46 am
by EViews Gareth
Depends on how you specified the equation.
Re: log of the intercept in least squares
Posted: Fri Jul 30, 2010 10:01 am
by trout
The equation was specified as: log(y)=log(C(1))-C(2)*log(x). I didn't realize before but it has "Convergence achieved after 6 iterations", so it must be nls. Which makes sense seeing as though it's nonlinear in the coefficients.
Re: log of the intercept in least squares
Posted: Fri Jul 30, 2010 10:12 am
by EViews Gareth
Indeed.