Scalable Estimation of Multinomial Response Models with Random Consideration Sets

Siddhartha Chib,Kenichi Shimizu

DOI: https://doi.org/10.48550/arXiv.2308.12470

2024-09-01

Abstract:A common assumption in the fitting of unordered multinomial response models for $J$ mutually exclusive categories is that the responses arise from the same set of $J$ categories across subjects. However, when responses measure a choice made by the subject, it is more appropriate to condition the distribution of multinomial responses on a subject-specific consideration set, drawn from the power set of $\{1,2,\ldots,J\}$. This leads to a mixture of multinomial response models governed by a probability distribution over the $J^{\ast} = 2^J -1$ consideration sets. We introduce a novel method for estimating such generalized multinomial response models based on the fundamental result that any mass distribution over $J^{\ast}$ consideration sets can be represented as a mixture of products of $J$ component-specific inclusion-exclusion probabilities. Moreover, under time-invariant consideration sets, the conditional posterior distribution of consideration sets is sparse. These features enable a scalable MCMC algorithm for sampling the posterior distribution of parameters, random effects, and consideration sets. Under regularity conditions, the posterior distributions of the marginal response probabilities and the model parameters satisfy consistency. The methodology is demonstrated in a longitudinal data set on weekly cereal purchases that cover $J = 101$ brands, a dimension substantially beyond the reach of existing methods.

Methodology,Econometrics,Applications,Computation

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to deal with the reality that different individuals have different consideration sets in the multi - category response model. Specifically, traditional multi - category response models assume that all individuals' choices come from the same set of categories, but this assumption may not hold in practical applications, especially in fields such as economics and marketing, where individuals may only choose from some of the available options. If this heterogeneity of choice is ignored, it will lead to bias in model parameter estimation, which in turn will affect the understanding of covariate effects and decision - making. To solve this problem, the author proposes a new method for estimating generalized multi - category response models. The core of this method is that it allows the response distribution of each individual to be conditional on a specific consideration set, which is randomly drawn from all possible consideration sets. This method can more accurately reflect individual choice behavior, and by introducing a method of mixing independent consideration models, it enables effective parameter estimation even when the number of categories is large. In addition, the author also proves that under certain conditions, the posterior distribution of model parameters is consistent, thus providing theoretical support for the effectiveness of the model. In short, the main contribution of this paper is to provide a new method that can both handle the complexity of consideration set heterogeneity and achieve efficient computation on high - dimensional data, which is of great significance for improving the accuracy and reliability of multi - category response models in practical applications.

Scalable Estimation of Multinomial Response Models with Random Consideration Sets

Multiple Imputation of Missing Categorical and Continuous Values via Bayesian Mixture Models with Local Dependence

Bayesian estimation and prediction for certain mixtures

Scalable Bayesian Estimation in the Multinomial Probit Model

Structured Mixture of Continuation-ratio Logits Models for Ordinal Regression

Stochastic variational inference for large-scale discrete choice models using adaptive batch sizes

Model selection in finite mixture of regression models: a Bayesian approach with innovative weighted g priors and reversible jump Markov chain Monte Carlo implementation (vol 85, pg 2456, 2015)

On the "Poisson Trick" and its Extensions for Fitting Multinomial Regression Models

Revealed Preference at Scale: Learning Personalized Preferences from Assortment Choices

Computationally Efficient Estimation of Large Probit Models

Joint Maximum Likelihood Estimation for High-dimensional Exploratory Item Response Analysis

On the Estimation of Discrete Choice Models to Capture Irrational Customer Behaviors

Inferential Tools for Assessing Dependence Across Response Categories in Multinomial Models with Discrete Random Effects

Bayesian mixture modeling for multivariate conditional distributions

Model based clustering of multinomial count data

A Nonparametric Approach with Marginals for Modeling Consumer Choice

Full-model estimation for non-parametric multivariate finite mixture models

Combinatorial Inference on the Optimal Assortment in Multinomial Logit Models

Flexible Modeling of Conditional Distributions Using Smooth Mixtures of Asymmetric Student T Densities

Bayesian Analysis of Binary and Polychotomous Response Data

A Random-effects Approach to Regression Involving Many Categorical Predictors and Their Interactions