One-sided tests for linear panel data models

This paper proposes simple tests for the error component model for the joint null hypothesis of absence of random effects and positive first order serial correlation. These tests explicitely exploit the one-sided nature of the alternative hypotheses, which yields a power gain compared to existing pr...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autor principal: Sosa Escudero, Walter
Formato: Objeto de conferencia
Lenguaje:Inglés
Publicado: 2001
Materias:
Acceso en línea:http://sedici.unlp.edu.ar/handle/10915/173776
Aporte de:
Descripción
Sumario:This paper proposes simple tests for the error component model for the joint null hypothesis of absence of random effects and positive first order serial correlation. These tests explicitely exploit the one-sided nature of the alternative hypotheses, which yields a power gain compared to existing procedures based on two-sided alternatives. The test statistics are, by construction, asymptotically optimal. A Monte Carlo experiment shows that the proposed statistics have good size and power performance in very small samples like those typically used in applied work in panel data.