Project R-3159

Statistical methods for the analysis of surveys with an informative spatio-temporal sampling design. (Research)

Investigation of spatio-temporal trends involves the fitting of a spatial and temporal process to sampled subjects at different geographical locations. Preferential sampling arises when the process that determines the data locations and the process being modeled are stochastically dependent. Conventional statistical methods do not account for the sampling design and assume that the sampling is non-preferential. This however, can lead to seriously biased estimates. In this project, methods will be developed that account for the informative sampling design based on jointly analyzing the sampling- and data-process and based on weighted versions of the data-process distribution. To understand how the sampling design effects the inference on parameters of interest, the ignorability of the spatio-temporal sampling design with respect to the inferential framework will be investigated. Also, a test for ignorability will be proposed in order to know when we can ignore the sampling design and restrict attention to the conventional statistical methods. This project will study the statistical theoretical properties of spatio-temporal sampling design, and will illustrate the studied methodology using different case studies.

01 October 2011 - 30 September 2015