Events and Meetings of Italian Statistical Society, Advances in Latent Variables - Methods, Models and Applications

Font Size: 
Approximate likelihood inference via dimension reduction for multidimensional binary panel data
Silvia Bianconcini, Silvia Cagnone, Dimitris Rizopoulos

Last modified: 2013-06-16

Abstract


Latent variable models represent a useful tool for the analysis of complex data when the constructs of interest are not directly observable. One problem related to these models is that the integrals involved in the maximization of the likelihood function cannot be solved analytically. Several solutions have been proposed in the literature, such as the Gauss-Hermite (GH) and the adaptive GH quadratures, but they become unfeasible in presence of multidimensional panel data, in which the number of latent variables directly increases with the number of observed items. To overcome these limitations, in this paper we propose a new approach for approximating integrals in latent variable models that has been introduced by Rahman and Xu (2004) in the engineering literature. It consists of reducing the dimensionality of the integrals involved in the computations.

Full Text: PDF