Extending the logit-mixed logit model for a combination of random and fixed parameters

The logit-mixed logit (LML) model, which allows the analyst to semi-parametrically specify the mixing distribution of preference heterogeneity, is a very recent advancement in logit-type choice models. In addition to generalize many previous semi- and non-parametric specifications, LML is computationally very efficient due to a computationally-convenient likelihood equation that does not require computation of choice probabilities in iterative optimization. However, the original LML formulation assumes all utility parameters to be random. This study extends LML to a combination of fixed and random parameters (LML-FR), and motivates such combination in random parameter choice models in general. We further show that the likelihood of the LML-FR specification loses its special properties, leading to a much higher estimation time. In an empirical application about preferences for alternative fuel vehicles in China, estimation time increased by a factor of 20–40 when introducing fixed parameters. Despite losses in computation efficiency, we show that the flexibility of LML-FR is essential for retrieving eventual multimodality of mixing distributions.

[1]  Georg Bühler,et al.  The impact of fuel availability on demand for alternative-fuel vehicles , 2012 .

[2]  Stephane Hess,et al.  Competing methods for representing random taste heterogeneity in discrete choice models , 2006 .

[3]  D. McFadden.  Conditional logit analysis of qualitative choice behavior , 1972 .

[4]  Gary J. Russell,et al.  A Probabilistic Choice Model for Market Segmentation and Elasticity Structure , 1989 .

[5]  Stefan Lindhard Mabit,et al.  Easy and flexible mixture distributions , 2013 .

[6]  M. Fosgerau,et al.  Easy and exible mixture distributions , 2013 .

[7]  Mark J. Garratt,et al.  Efficient Experimental Design with Marketing Research Applications , 1994 .

[8]  Ryuichi Kitamura,et al.  Demand for clean-fuel vehicles in California: A discrete-choice stated preference pilot project , 1993 .

[9]  Kenneth Train,et al.  Mixed logit with a flexible mixing distribution , 2016 .

[10]  Jj Louviere,et al.  Confound it! That pesky little scale constant messes up our convenient assumptions , 2006 .

[11]  Jeremy T. Fox,et al.  Linear Regression Estimation of Discrete Choice Models with Nonparametric Distributions of Random Coefficients , 2007 .

[12]  D. McFadden,et al.  MIXED MNL MODELS FOR DISCRETE RESPONSE , 2000 .

[13]  Michael C Sachs,et al.  plotROC: A Tool for Plotting ROC Curves. , 2017, Journal of statistical software.

[14]  Philippe L. Toint,et al.  Estimating Nonparametric Random Utility Models with an Application to the Value of Time in Heterogeneous Populations , 2010, Transp. Sci..

[15]  Ricardo A. Daziano,et al.  Multinomial Logit Models with Continuous and Discrete Individual Heterogeneity in R: The gmnl Package , 2017 .

[16]  Kenneth Train,et al.  EM algorithms for nonparametric estimation of mixing distributions , 2008 .

[17]  Michel Bierlaire,et al.  A practical test for the choice of mixing distribution in discrete choice models , 2005 .

[18]  J. Boyd,et al.  THE EFFECT OF FUEL ECONOMY STANDARDS ON THE U.S. AUTOMOTIVE MARKET: AN HEDONIC DEMAND ANALYSIS , 1980 .

[19]  Jeremy T. Fox,et al.  A simple estimator for the distribution of random coefficients , 2011 .

[20]  Jordan J. Louviere,et al.  Formal Choice Models of Informal Choices , 2008 .

[21]  Chandra R. Bhat,et al.  An Endogenous Segmentation Mode Choice Model with an Application to Intercity Travel , 1997, Transp. Sci..

[22]  Steven H. Cohen,et al.  Market segmentation with choice-based conjoint analysis , 1995 .