The analysis of non-metric endogenous variables in latent variable models: The MECOSA approach

Abstract

In this paper we present models for the analysis of microeconomic data that can handle limited dependent variables (metric, censored metric, dichotomous, ordinal) simultaneously. In these models the dependent variables are indicators of latent variables that are the endogenous variables in a simultaneous equation system. Such models may be considered as generalizations of models with errors in the variables. We also introduce a new computer program (Mecosa: MEan and COvariance Structure Analysis) for estimating the model parameters. Unlike the Liscomp program (Muthén 1988), Mecosa can treat the case where the latent variables are hierarchically structured, and Mecosa can also handle more general restrictions on the parameters, subject only to the condition that each parameter can be expressed as a continuously differentiable function of the set of fundamental parameters. The program itself is implemented in Gauss (Version 2.0) for IBM-PC’s and compatibles.KeywordsFactor analysishierarchical structures Lisrel, Liscomp polychoric correlation coefficientprobit modeltobit model.