Comparação de métodos para teste de bigaussianidade

In geostatistics some methods work under the strong hypothesis of multiGaussianity. Sequential Gaussian simulation and multiGaussian kriging are examples of these methods that require original data transformed into normal score. However, it does not guarantee that cumulative distributions functions for two, three or more points follow a Gaussian distribution as well. All available tests so far have used two-point statistics that is best described by a semivariogram function. If the test for two points is positive then the multiGaussian model can be adopted. In this paper two methods are tested with three samples drawn from an exhaustive data set. Results of these methods are compared each other and in conclusion one of them can be considered statistically robust.