Testing Mixed Logit and Probit Models by Simulation

2005 ◽  
Vol 1921 (1) ◽  
pp. 53-62 ◽  
Author(s):  
Marcela A. Munizaga ◽  
Ricardo Alvarez-Daziano
Keyword(s):  
Discrete Choice ◽  
Recent Literature ◽  
Scale Effects ◽  
Flexible Structures ◽  
Mixed Logit ◽  
Discrete Choice Models ◽  
Modeling Tools ◽  
Probit Models ◽  
Logit And Probit Models ◽  
Halton Sequences

Discrete choice models with error structures that are not independent and identically distributed have received enormous attention in the recent literature. A detailed synthetic study tests this type of model in a controlled case. With mixed logit and probit models as the study objects, calibration was implemented with the use of software available on the Internet. The controlled situation was built as a simulation laboratory, which generated databases with known parameters. The effects of various elements were analyzed: number of repetitions of the simulation, number of observations in the database, and how the use of Halton sequences improves the mixed logit calibration. The scale effects on the different models are also discussed. The results obtained in this specific context lead to some recommendations for future users of these powerful modeling tools. In particular, flexible structures require large sample sizes to calibrate the elements of the covariance matrix.

Analyzing Bus Passengers’ Satisfaction in Dhaka using Discrete Choice Models

2019 ◽  
Vol 2673 (2) ◽  
pp. 758-768 ◽  
Author(s):  
Mohammed Quddus ◽  
Farzana Rahman ◽  
Fredrik Monsuur ◽  
Juan de Ona ◽  
Marcus Enoch
Keyword(s):  
Service Quality ◽  
Influencing Factors ◽  
Transport System ◽  
Discrete Choice ◽  
Logit Model ◽  
Mixed Logit ◽  
Multinomial Logit ◽  
Discrete Choice Models ◽  
Bus Service ◽  
Bus Transport

The bus transport system in Dhaka is unsafe, unreliable, inefficient and struggles to cope with the day-to-day mobility of its massive population. Consequently, measuring the performance of bus service quality (SQ) from the customers’ perspective is fundamental in planning a sustainable bus transport system for Dhaka, and in developing the associated policies and regulations. Although there are some studies addressing the performance of the public transport systems in Bangladesh, little research considers how service quality attributes affect passengers’ satisfaction. The purpose of this paper is to examine a relationship between bus service quality and its influencing factors in Dhaka. Using a customer satisfaction survey with a sample size of 955, discrete choice models (e.g., multinomial logit and mixed logit) have been developed. The results indicate that the inhabitants, as expected, are dissatisfied with their bus services (less than 10% rated service quality as “excellent/good”) and service attributes such as comfort level and driver skills were found to be the most important contributors toward the “poor” and “very poor” perceptions of service quality. Other influencing factors are punctuality, safety, entry and exit processes, waiting times, and vehicle condition. One surprising finding was that the multinomial logit model provides better goodness-of-fit for the sample data relative to the mixed logit model implying that bus users in Dhaka may represent a homogeneous group as they do have access to other modes. Findings from this study can be utilized to develop policies and regulations to improve bus transport in Dhaka.

scholarly journals Performance Analysis of Mixed Logit Models for Discrete Choice Models

2019 ◽  
pp. 563-575
Author(s):  
Jaka Nugraha
Keyword(s):  
Discrete Choice ◽  
Logit Model ◽  
Mixed Logit ◽  
Likelihood Estimation ◽  
Multinomial Logit Model ◽  
Discrete Choice Models ◽  
Logit Models ◽  
Nominal Data ◽  
Mixed Logit Models ◽  
Mnl Model

Mixed Logit model  (MXL) is generated from Multinomial Logit model (MNL) for discrete, i.e. nominal, data. It eliminates its limitations particularly on estimating the correlation among responses.  In the MNL, the probability equations are presented in the closed form and it is contrary with in the MXL. Consequently, the calculation of the probability value of each alternative get simpler in the MNL, meanwhile it needs the numerical methods for estimation in the MXL.  In this study, we investigated the performance of maximum likelihood estimation (MLE) in the MXL and MNL into two cases, the low and high correlation circumstances among responses. The performance is measured based on differencing actual and estimation value.  The simulation study and real cases show that the MXL model is more accurate than the MNL model. This model can estimates the correlation among response as well. The study concludes that the MXL model is suggested to be used if there is a high correlation among responses. 

Discrete Choice Methods

2009 ◽  
Author(s):  
Garrett Glasgow ◽  
R. Michael Alvarez
Keyword(s):  
Political Science ◽  
Discrete Choice ◽  
Statistical Models ◽  
Mixed Logit ◽  
Science Research ◽  
Choice Models ◽  
Discrete Choice Models ◽  
Choice Situation ◽  
Multinomial Probit ◽  
Political Agents

This article describes the statistical models commonly used to study discrete choices. It concentrates on the ‘basic’ discrete choice models, and the theoretical choice situations that lead to these models. Specifically the choice situation addressed include: the ordered choice situation and the unordered choice situation. In addition, the article discusses two extensions of the basic discrete choice models commonly seen in political science research — models allowing for heteroskedasticity in the choices made across political agents (such as the heteroskedastic probit), and models that estimate substitution patterns across choice alternatives (such as the multinomial probit and mixed logit). Suggestions for further reading are also given.

Exploring Marketing to Engineering Information Mapping in Mass Customization: A Presentation of Ideas, Challenges and Resulting Questions

2011 ◽  
Author(s):  
Scott Ferguson ◽  
Andrew Olewnik ◽  
Phil Cormier
Keyword(s):  
Discrete Choice ◽  
Mass Customization ◽  
Latent Class ◽  
Mixed Logit ◽  
Choice Models ◽  
Discrete Choice Models ◽  
Hierarchical Bayes ◽  
Customer Preferences ◽  
Lack Of Information ◽  
Engineering Information

The paradigm of mass customization strives to minimize the tradeoffs between an ‘ideal’ product and products that are currently available. However, the lack of information relation mechanisms that connect the domains of marketing, engineering, and distribution have caused significant challenges when designing products for mass customization. For example, the bridge connecting the marketing and engineering domains is complicated by the lack of proven tools and methodologies that allow customer needs and preferences to be understood at a level discrete enough to support true mass customization. Discrete choice models have recently gained significant attention in engineering design literature as a way of expressing customer preferences. This paper explores how information from choice-based conjoint surveys might be used to assist the development of a mass customizable MP3 player, starting from 140 student surveys. The authors investigate the challenges of fielding discrete choice surveys for the purpose of mass customization, and explore how hierarchical Bayes mixed logit and latent class multinomial logit models might be used to understand the market for customizable attributes. The potential of using discrete choice models as a foundation for mathematically formulating mass customization problems is evaluated through an investigation of strengths and limitations.

scholarly journals Modeling enterprise location choice decision behavior

2021 ◽  
Vol 14 (1) ◽  
pp. 669-691
Author(s):  
Nguyen Cao Y
Keyword(s):  
Spatial Correlation ◽  
Discrete Choice ◽  
Choice Model ◽  
Location Choice ◽  
Mixed Logit ◽  
Multinomial Logit ◽  
Choice Models ◽  
Discrete Choice Models ◽  
Spatial Effects ◽  
Modeling Framework

This study presents a location choice model that incorporates urban spatial effects for enterprises. A modeling framework is developed to analyze decisions regarding location choice for enterprises using a series of discrete choice models including multinomial logit without any urban spatial effects, multinomial logit incorporating urban spatial effects, and mixed logit incorporating urban spatial effects. In this framework, urban spatial effects, such as the urban spatial correlation among enterprises in deterministic terms and the urban spatial correlation among zones in the error term, are captured by mixed logit models in particular and discrete choice models in general. The results indicate that the urban spatial effects and the land prices in a given zone strongly affect the decision-making process of all the enterprises in the Tokyo metropolitan area. Moreover, the important role of urban spatial effects in the proposed model will be clarification through comparing the three above models. This comparison will be implemented on the basis of three types of indicators such as the log likelihood ratio, Akaike information indicator, and hit ratio of each model.

Comparative analysis of the spatial distribution of fishing effort contrasting ecological isodars and discrete choice models

2014 ◽  
Vol 71 (1) ◽  
pp. 141-150 ◽  
Author(s):  
A. van der Lee ◽  
D.M. Gillis ◽  
P. Comeau
Keyword(s):  
Discrete Choice ◽  
Choice Model ◽  
Ideal Free Distribution ◽  
Mixed Logit ◽  
Spatial Dynamics ◽  
Predictive Performance ◽  
Discrete Choice Models ◽  
Fishing Effort ◽  
Otter Trawl ◽  
Random Utility

The spatial dynamics of catch and effort data are often overlooked in fisheries research despite its well-documented utility in understanding the distribution and abundance of fish. We apply a recently developed fisheries isodar model to an otter trawl groundfish fishery on the Scotian Shelf and compare its predictive performance with a more traditional discrete choice model random utility model. Isodars represent the expected distribution of foragers between two habitats when fitness is equal and can be a representation of the ideal free distribution. Here, fitness was defined with relative catch rates, cost differentials, and interference effects between habitats. Random utility models were constructed as mixed logit models to give the expected probability of fishing in a particular area based on a collection of predictors. The predictions of the isodar models consistently outperformed the mixed logit for both in-sample and out-of-sample forecasts and the isodar was determined to be the preferred model based on its increased accuracy and simplicity. The isodar model can provide a statistically powerful and easily implemented tool in effort studies, especially in situations of aggregated or limited data, which can inform conservation and management decisions.

Sign in / Sign up

Export Citation Format

Share Document