Estudio de Monte Carlo para comparar 8 pruebas de normalidad sobre residuos de mínimos cuadrados ordinarios en presencia de procesos autorregresivos de primer orden.

Abstract

The objective of this study is to assess the statistical power and size of 8 normality tests in presence of first-order autoregressive errors and different simple sizes. Using aMonte Carlo experiment, the following tests were compared: Cramér-von Mises, Anderson-Darling, Lilliefors, Pearson, Shapiro-Wilk, ShapiroFrancia, Jarque-Bera and D’Agostino-Pearson. Our results show 4 relevant findings: First, an asymmetrical effect of autocorrelation on the power and size of the tests. Second, the statistical size of all tests is affected by the autocorrelation. Third, none of the tests has greater power than the others. Fourth, the power of the normality test decreases as sample size decreases.