Sensitivity analysis of the Jarque-Bera
and Lilliefors normality tests in
linear regression models

Marcelo Rodrigo Portela FERREIRA[1]

§    ABSTRACT: One of the assumptions regarding the linear regression model is that errors are assumed to follow normal distribution. This is not always necessary; however, most of the times it is reasonable to consider it because the errors normality assumption becomes important when we want to construct confidence intervals and conduct hypotheses tests. There are some tests that are used to assess if one given univariate sample follows normal distribution or not. In this article, we will evaluate, through Monte Carlo Simulation, the performance under diverse scenarios of two tests that are probably the most widely used: the Jarque-Bera test and the Lilliefors test.

§    KEYWORDS: Normality; Jarque-Bera test; Lilliefors test; Monte Carlo simulation.



[1] Departamento de Estatística, Centro de Ciências Exatas e da Natureza, Universidade Federal de Pernambuco -–UFPE, CEP: 50740-540, Recife, PE, Brasil, E-mail: marcelorpf@gmail.com