A Note on Exact Computation of Max Weighted Score Estimators by Mixed Integer Programming

22 Pages Posted: 16 Aug 2007 Last revised: 8 Dec 2008

See all articles by Kostas Florios

Kostas Florios

Athens University of Economics and Business

Spyros Skouras

Athens University of Economics and Business - Department of International and European Economic Studies

Date Written: August 1, 2007

Abstract

We show that exact computation of a family of 'max weighted score' estimators, including Manski's max score estimator, can be achieved efficiently by reformulating them as mixed integer programs (MIP) with disjunctive constraints. The advantage of our MIP formulation is that estimates are exact and can be computed using widely available solvers in reasonable time. In a classic work-trip mode choice application our method delivers exact estimates that lead to a different economic interpretation of the data than previous heuristic estimates. In a small Monte Carlo study we find that our approach is computationally efficient for usual estimation problem sizes.

Keywords: maximum score, mixed integer programming, estimator computation, work-trip mode choice

JEL Classification: C25, C61, C63 , C87

Suggested Citation

Florios, Kostas and Skouras, Spyros, A Note on Exact Computation of Max Weighted Score Estimators by Mixed Integer Programming (August 1, 2007). Available at SSRN: https://ssrn.com/abstract=1007441 or http://dx.doi.org/10.2139/ssrn.1007441

Kostas Florios

Athens University of Economics and Business ( email )

HOME PAGE: http://sites.google.com/site/kflorios/

Spyros Skouras (Contact Author)

Athens University of Economics and Business - Department of International and European Economic Studies ( email )

GR-10434 Athens
Greece

Here is the Coronavirus
related research on SSRN

Paper statistics

Downloads
257
Abstract Views
1,681
rank
130,513
PlumX Metrics