Analyzing Multilevel Spatially Correlated Data Using Composite Quadratic Inference Functions

• Activity Number: 42
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Yun Bai*+ and Peter Song
• Companies: University of Michigan and University of Michigan
• Keywords: Spatial data ; multilevel ; quadratic inference function ; composite likelihood

Abstract

Spatially correlated data are frequently encountered in epidemiology studies. Data structure is further complicated by the presence of hierarchy, i.e. subjects are nested within geographic units which are spatially correlated. To deal with the high dimensionality of the data, we apply the composite quadratic inference function approach to account for between- and within-group correlations. The idea is to reduce dimension by building two sets of estimating equations based on within- and between-group pairs separately and then combine them through a weight matrix in the form of the quadratic inference function to improve efficiency. We conduct simulation experiments to assess the performance of our method for Gaussian and binary outcomes. The method is then applied to analyze the prevalence of malaria among a sample of village resident children in Gambia.