Proposta de um teste de hipótese para a existência de dependência espacial em dados geoestatísticos

Abstract

The assessment of the significance of spatial dependence is important for the formal inference about the null hypothesis of non-existence of spatial dependence in geostatistics. Through this study, we aimed at developing a significance test for the null hypothesis of lack of spatial dependence, for the best decision about the existence whether or not the dependence on geostatistical data. For the development of the spatial dependence test, we considered the characteristics of models adjusted to the semivariogram; and for the development of the hypothesis test, firstly the null hypothesis and a statistics of the test generated from the concept of spatial dependence area obtained in the semivariogram were defined. The test was developed based on simulation of geostatistical phenomena presenting features of pure nugget effect, i.e., phenomena in which the null hypothesis was true. Finally, it was studied the power of the test for different degrees of simulated spatial dependencies. The test showed good power, and this tended to 100% when the degree of spatial dependence of the simulated geostatistical phenomena increased.