lclogit2: An Enhanced Module to Estimate Latent Class Conditional Logit Models
20 Pages Posted: 25 Nov 2019
Date Written: November 10, 2019
Abstract
This paper describes Stata command lclogit2, an enhanced version of lclogit (Pacifico and Yoo, 2013). Like its predecessor, lclogit2 uses the Expectation-Maximization (EM) algorithm to estimate latent class conditional logit (LCL) models. But it executes the EM algorithm's core algebraic operations in Mata, and runs considerably faster as a result. It also allows linear constraints on parameters to be imposed in a more convenient and flexible manner. It comes with parallel command lclogitml2, a new standalone program that uses gradient-based algorithms to estimate LCL models. Both lclogit2 and lclogitml2 are supported by a new postestimation tool, lclogitwtp2, that evaluates willingness-to-pay measures implied by estimated LCL models.
Note: As of 10 November 2019, this paper is under review at the Stata Journal.
Keywords: lclogit2, lclogitml2, lclogitwtp2, lclogit, mixlogit, fmm, finite mixture, mixed logit
JEL Classification: C35, C61, C87
Suggested Citation: Suggested Citation