A Latent Class Nested Logit Model for Rank-Ordered Data with Application to Cork Oak Reforestation

lclogit2: An enhanced command to fit latent class conditional logit models

“…In the context of a finite or discrete mixture logit model, which LCL is, whether users fit the model in one space or another is less critical. As Oviedo and Yoo (2017) point out, the set of mass points in a discrete mixing distribution that maximizes the 20. Yoo and Doiron (2013) and Doiron and Yoo (2020) provide further information on identification and interpretation of the max-diff logit models' coefficients and their comparisons with the traditional clogit utility coefficients.…”

“…Often, the pecuniary attribute measures the cost of acquiring a particular alternative. For example, in Oviedo and Yoo (2017), each alternative is a reforestation project, and the cost is a required increase in the decision maker’s tax liabilities to finance that project. In some applications, the pecuniary attribute may measure income generated by a particular alternative instead.…”

“…How many classes, C , should LCL allow for? In many empirical studies, including my own (Yoo and Doiron 2013; Doiron and Yoo 2017, 2020; Oviedo and Yoo 2017), this question is addressed by repeatedly fitting the same LCL model with different numbers of classes and inspecting which number leads to the best model in terms of the BIC. calculates and stores the fitted model’s BIC in , facilitating this specification search.…”

“…In the context of a finite or discrete mixture logit model, which LCL is, whether users fit the model in one space or another is less critical. As Oviedo and Yoo (2017) point out, the set of mass points in a discrete mixing distribution that maximizes the sample log-likelihood function is invariant to whether the model is parameterized in the preference space or the WTP space. Therefore, the WTP measures derived from the utility coefficients (using ) are the same as what users would have obtained if they reparameterized the model to fit the WTP measures directly.…”

“…Fourth, lclogitml2 is a stand-alone estimation command. lclogitml, which accompanies lclogit, is simply a wrapper that passes lclogit estimates to another community-contributed command, gllamm (Rabe-Hesketh, Skrondal, and Pickles 2002). This difference brings about several advantages: a. lclogitml2 uses a log-likelihood evaluator coded in Mata.…”