UJI KENORMALAN UNIVARIAT: SUATU KAJIAN PUSTAKA

Abstract

Almost all statistical procedures, especially statistical inference, assumed that the sample distribution is
normally distributed. This normality assumption must be tested to ensure the correct use of the test
statistic, hence resulting a correct conclusion. This research discusses some univariate normality tests:
test based on empirical distribution function, test based on moments, test based on correlation or
regression, test based on sample entropy, test based on kernel method, test based on Polya
characteristics, and test based on nonparametric method. This research also discuss normality test that
capable of detecting outliers and discuss omnibus test that can give additional information about nonnormality.