A Note on the Estimation of the Multinomial Logit Model With Random Effects

• Published in: The American statistician Vol. 55; no. 2; pp. 89 - 95
• Main Authors: Chen, Zhen, Kuo, Lynn
• Format: Journal Article
• Language: English
• Published: Alexandria, VA Taylor & Francis 01.05.2001; American Statistical Association
• Subjects: Brands; Censorship; Datasets; Discrete choice model; Economic models; Economic statistics; Estimation methods; Exact sciences and technology; Failure analysis; Generalized linear models; Linear inference, regression; Marketing; Mathematical models; Mathematics; Matrices; Modeling; Multilevel models; Multinomial logit model; Observed choices; Poisson log-linear model; Poisson nonlinear model; Polytomous data; Probability and statistics; Sciences and techniques of general use; Software; Statistical Practice; Statistics; Survival analysis; Unobserved heterogeneity; Yogurt; Non linear regression; Fitting; Loglinear model; Discrete data; Statistical estimation; Poisson distribution; Algorithm; Implementation; Linear model; Statistical regression; Heterogeneity; Equivalence; Computer program; Correlation analysis; Non linear model; Random effect; Multinomial distribution; Linear estimation; Logit model; Likelihood function; Proportional hazard model
• ISSN: 0003-1305; 1537-2731
• DOI: 10.1198/000313001750358545

The multinomial logit model with random effects is often used in modeling correlated nominal polytomous data. Given that there is no standard software of fitting it, we advocate using either a Poisson log-linear model or a Poisson nonlinear model, both with random effects. Their implementations can be carried out easily by many existing commercial statistical packages including SAS. A brand choice dataset is used to illustrate the proposed methods.