Testes Bootstrap para detecção de anisotropia espacial em fenômenos contínuos

Abstract

Anisotropy is a trait in which spatially dependent data show different behaviors in certain directions. When the behavior of the random variable is similar in all directions, we say that the phenomenon is isotropic. In case anisotropy is not detected and, consequently, its correlation is not made, the phenomenon can be wrongly interpreted. But the lack of a consolidated statistical test for detecting anisotropy corroborated the difficulty the researcher faces to correct this characteristic of the phenomenon. This work proposes two tests to detect spatial anisotropy: one based on the area and another based on Euclidean distances. To verify the efficacy of the tests, we computed the Type I error rates and power of the test. In all simulated configurations, the test based on the area obtained better performance than the test based on Euclidean distances.