R. Giraldo, R. Giraldo
Methods are proposed to carry out spatial prediction of curves when a sample of curves taken at different sites in a region is available. Predictors called ordinary kriging, pointwise, and total functional kriging are presented. Alternatives are also shown for the case of nonstationary functional random fields. The functional residual kriging and universal kriging predictors are considered in this scenario. The predictors have the same principle as the classic kriging used in univariate geostatistics; that is, the weights are obtained by minimizing the prediction variance, taking into account unbiasedness restrictions (best unbiased linear predictor). As part of the proposal, the concept of variogram used in univariate geostatistics to estimate the spatial correlation structure of the random field under observation is extended to the case of functional data. The methodologies are illustrated through the analysis of real data.
Keywords: Functional data, functional stationary and non stationary processess, ordinary kriging, universal kriging, residual kriging
Scheduled
Invited Session Colombian Society of Statistics. Spatial Statistics for functional data
June 7, 2022 6:40 PM
Conference hall