EcoSta 2018: Registration
View Submission - EcoSta2018
A0225
Title: Composite likelihood inference for replications of spatial ordinal data Authors:  Jun Zhu - University of Wisconsin (United States)
Pingping Wang - Nanjing University of Finance and Economics (China) [presenting]
Abstract: Spatial ordinal data observed on multiple subjects are common in practice yet statistical methodology for such ordinal data analysis is limited. The existing methodology often assumes a single realization of spatial ordinal data without replications and thus, it is not directly applicable for the kind of spatial ordinal data observed on multiple subjects. We develop a spatial ordinal probit model that enables the assessment of covariates via regression and accounts for spatial correlation via a geostatistical model. We then develop maximum composite likelihood method for parameter estimation and establish the asymptotic properties of the parameter estimates, which differs from the existing literature in that the number of subjects tends to infinity but not the number of spatial locations per subject. The asymptotic properties permit an approximate estimation of the variance of the parameter estimates and facilitate the inference for the model parameters in a computationally efficient manner. A simulation study suggests sound finite sample properties of the proposed methods and a real data example in dental health is presented for illustration.