Evaluating three proposals for testing independence in non linear spatial processes

  1. López Hernández, Fernando Antonio
  2. Maté Sánchez de Val, Mari Luz
  3. Artal Tur, Andrés
Revista:
Estadística española

ISSN: 0014-1151

Ano de publicación: 2013

Volume: 55

Número: 180

Páxinas: 55-76

Tipo: Artigo

Outras publicacións en: Estadística española

Resumo

This paper evaluates the behaviour of different families of tests when checking for spatial independence in the presence of nonlinearities. To reach this goal, we select three representative proposals. The usual parametric tests of I-Moran, the nonparametric proposal of Brett and Pinkse (1997), and the semiparametric Scan test. In order to study how they perform, we simulate different nonlinear spatial structures by Monte Carlo methods, hence conducting empirical tests on the matter. Main results show failures of traditional tests in this framework, and the need to build on new proposals in the presence of nonlinearities. An empirical application to an economic-theory-of-production scenario illustrates the performance of the three tests.