CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: Pseudo-population bootstrap for design-based inference on spatial phenomena Authors:  Lorenzo Fattorini - University of Siena (Italy) [presenting]
Abstract: Spatial prediction on continuous surfaces or for finite populations of points or spatial units are usually performed in a model-dependent framework, assuming spatial processes generating the values of the survey variable for depicting his map on the whole study area. Recently, the properties of inverse distance weighting interpolation were derived in a completely design-based framework. Asymptotic scenarios and conditions ensuring design-based asymptotic unbiasedness and consistency were derived. They mainly require smoothness of the variable under study and the use of spatially balanced sampling schemes. As the resulting map constitutes a pseudo-population converging to the true spatial population, it can be adopted in a resampling procedure, selecting bootstrap samples from it by means of the spatial scheme actually adopted to select the sample from which interpolation has been performed. The procedure can be used to make inference on the distribution of complex estimators of the spatial population parameters as well as to make inference on the estimated map.