CMStatistics 2018: Start Registration
View Submission - CMStatistics
B1190
Title: Small area estimation in the framework of multivariate models for sustainable development Authors:  Emilia Rocco - University of Florence (Italy)
Maria Francesca Marino - University of Florence (Italy)
Alessandra Petrucci - University of Firenze (Italy)
Emilia Rocco - University of Florence (Italy) [presenting]
Abstract: The analysis of complex phenomena, such as the equitable and sustainable development, often requires the estimation of correlated descriptive measures. Multivariate models are specifically designed to take into account the correlation of several variables and, typically, fit to these kind of situations. We introduce a multivariate mixed model for the case in which small area estimates of different, possibly heterogeneous, survey variables are required. In this framework, estimates of model parameters in the different equations may borrow strength one from the other and, thus, efficiency may be improved. In particular, this is achieved by introducing in the model specification a set of correlated latent effects that allow us to capture the dependence among the different outcomes of interest. Further, the estimated correlation between the latent effects provides an indirect measure of the dependence between the outcomes themselves and, thus, offers a deeper understanding of the phenomenon under investigation. Estimation of model parameters is discussed within a likelihood-based approach. Predictions of small area parameters are derived and a parametric bootstrap method is proposed for estimating the mean squared error of such predictions. Results are supported by an intensive simulation study in which different scenarios are investigated.