Combining Discrete and Continuous Representations of Preference Heterogeneity: A Latent Class Approach

Abstract

Travel cost method, Recreation demand, Random parameter model, Latent class model, Forests,

Random taste heterogeneity in discrete choice models: Flexible nonparametric finite mixture distributions

“…Like nonparametric distributions, finite mixtures of parametric distributions too can asymptotically mimic any shape, but high computational costs often preclude the estimation of models with a large number of mixtures, and models estimated in practice are typically less flexible than those with nonparametric distributions. For example, Fosgerau and Hess (2009) and Bujosa et al (2010) use discrete mixtures of normal distributions, and Greene and Hensher (2013) use discrete mixtures of triangular distributions, but each of these studies is limited by its attention solely to univariate distributions. Train (2008) expands the framework to allow for mixtures of multivariate normal distributions and their transformations, such as multivariate lognormal and truncated normal distributions, but the framework is empirically evaluated only for the case of mixtures of two independent multivariate distributions.…”

“…Since 2010 alone, we are aware of at least eleven new methods that have been proposed in the literature (c.f. Bastin et al, 2010;Bujosa et al, 2010;Fiebig et al, 2010;Bhat and Sidharthan, 2012;Bastani and Weeks, 2013;Greene and Hensher, 2013;Keane and Wasi, 2013;Dong and Koppelman, 2014;Train, 2016;Bansal et al, 2017;Bhat and Lavieri, 2017). However, for reasons expanded upon in the following section, each of these methods has proven inadequate in one way or another.…”

Adoption of renewable heating systems: An empirical test of the diffusion of innovation theory

“…For example, a new renewable fuel system may encroach more on options from a similar category of sustainable systems than on fossil fuel-based systems. To relax such a maintained assumption, we allowed for random taste variation within each class and estimated a Panel Latent Class-Random Parameters Logit model (LC-RPL) [57], [58], [59], [60], [27], [47], [61] accounting for the series of T choices made by each respondent.…”

“…Taste heterogeneity across households is therefore accounted for in two ways: (i) by identifying different behavioural classes as a function of the average score of the innovativeness scale and (ii) by considering continuous taste variation among individuals in the same group (within-group heterogeneity) [57].…”

Attribute processing as a behavioural strategy in choice making

“…Scarpa et al 2009 for example, find that over 90 percent of the sample ignore the cost attribute in the context of a stated preference survey designed to value landscapes in Ireland, where the cost attribute was specified as the value in Euros that the respondent would personally have to pay per year through their income tax and value added tax contributions. Greene and Hensher (2012), Bujosa et al (2010) and Hess et al (2011) introduce a natural extension of the fixed parameter latent class model as a random parameter latent class model which allows for another layer of preference heterogeneity within each class; however to date only , Hensher et al (2012a) and Collins et al (2012) have developed this model form in the context of AN-A. What we then have is a latent class model that allows for heterogeneity both within and across groups.…”