Box-Cox LogL convergence problem
Posted: Mon Dec 28, 2015 4:30 pm
Hello,
I'm having trouble estimating a Box-Cox model with 1+6 parameters (one for the dependent and a different one for each of the independent variables).
I have used all the possible combinations of the available optimization and step methods in the LogL object along with varying the convergence tolerance but it cannot converge.
I have also tried using numerical and/or analytic derivatives (they match the numeric calculations) to no avail.
Changing the initial values seems hopeless too, so far at least.
I was wondering if there are any suggestions on how to proceed.
thank you
ps. I have built on the eviews provided box-cox example with 1+1 parameters and it converged. I just can't make the multi-parameter version converge also.
ps2. I'm using eviews 9
I'm having trouble estimating a Box-Cox model with 1+6 parameters (one for the dependent and a different one for each of the independent variables).
I have used all the possible combinations of the available optimization and step methods in the LogL object along with varying the convergence tolerance but it cannot converge.
I have also tried using numerical and/or analytic derivatives (they match the numeric calculations) to no avail.
Changing the initial values seems hopeless too, so far at least.
I was wondering if there are any suggestions on how to proceed.
thank you
ps. I have built on the eviews provided box-cox example with 1+1 parameters and it converged. I just can't make the multi-parameter version converge also.
ps2. I'm using eviews 9