SOME LARGE SAMPLE TESTS FOR NONNORMALITY IN AR MODEL

Abstract

White and MacDonald ( 1980) proposed conditions under which several well-known and easily computable statistics for testing normality , ?b1, b2 , D* and W', modified for large sample use in autoregressive models by replacing the true stochastic error with the least squares residual. Monte Carlo simulations indicate that the modified statistics perform adequately well for moderate to large sample sized autoregressive models. Across distributions studied in this experiment, none of the tests clearly dominate over the others, although the modified W' performs well for moderate to large sample sizes. We illustrate the use of the normality tests using Box and Jenkin's study of Wolfer's Sunspot numbers.