ABSTRACT
Understanding the limitations of the sampling design adopted for each survey is clearly important in order to draw valid inferences from a joint analysis of the resulting data. In particular, non-randomized samples may reach an unrepresentative sub-population or be biased in other ways. Depending on the problem under investigation, the set of sampling locations might be the result of a natural process, for example the locations of nests in a colony of birds. The chapter shows an application to malaria prevalence data from a community survey and a school survey conducted in July 2010 in Rachuonyo South and Kisii Central Districts, Nyanza Province, Kenya. It considers the problem of carrying out geostatistical inference when the spatial locations X are recorded with error. The chapter focuses on geomasking which is obtained by adding a stochastic or deterministic displacement to the spatial coordinates X of a geostatistical data-set.
