[Limdep Nlogit List] Convergence problem in WTP-space estimation

[Robert Johnston]

David,

This is a common occurrence and is expected given differences between the two models.  When estimated in WTP-space, the utility function is nonlinear-in-the-parameters.  This leads to computational and convergence challenges beyond those encountered in an otherwise identical (linear-in-the-parameters) preference-space utility function.  It relatively common for WTP-space estimation to encounter local optima, flat spots in the likelihood surface, etc.  You might consider diagnosing your model using different starting values or convergence criteria, or try alternative optimization algorithms.  These can all be adjusted in NLOGIT.  Even under the best of conditions, WTP-space models sometimes require many iterations to converge.

Robert J. Johnston
Director, George Perkins Marsh Institute
Professor, Department of Economics
Clark University
950 Main St.
Worcester, MA 01610
Phone: (508) 751-4619
Fax: (508) 751-4600
Email: rjohnston at clarku.edu
https://protect-au.mimecast.com/s/z4nRBmUpn2pID?domain=clarku.edu


-----Original Message-----
From: Limdep [mailto:limdep-bounces at mailman.sydney.edu.au] On Behalf Of Ortega, David
Sent: Tuesday, August 15, 2017 10:29 AM
To: limdep at mailman.sydney.edu.au
Subject: [Limdep Nlogit List] Convergence problem in WTP-space estimation

Hello,

I am having trouble getting an RPL model to converge in WTP-space (the model reaches 500 iterations without converging). However, if I run the same RPL model in preference space and then run that model in WTP-space, the model converges. Can anyone provide insight as to what is causing this? This is happening in more than one dataset. I am using Nlogit5.0.

Thanks, David


_______________________________________________
Limdep site list
Limdep at mailman.sydney.edu.au
http://limdep.itls.usyd.edu.au