Review article
Distribution normality testing in the studies of education and upbringing
Siniša Opić
orcid.org/0000-0001-9800-0145
; Učiteljski fakultet Sveučilišta u Zagrebu
Abstract
Distribution normality testing, including the homogeneity of variance, is an inevitable parameter in the studies of education and upbringing due to the choice of certain statistic parametric or non-parametric tests for hypotheses testing. However, it has been unjustifiably excluded which can eventually cause distorted results (generalizations). The work describes statistic values of normal distribution and its modalities: kurtosis and skewness. It also describes other theoretical distributions which are often seen in the studies of education and upbringing: Poisson, uniform, t-distribution, hi-square and U-distribution. The tests for distribution normality have been as well elaborated: Kolmogorov-Smirnov test (K-S), Lilliefors test, Shapiro-Wilks test Anderson-Darling Normality Test.
As an empirical evaluation of statistically significant values of parametric and non-parametric statistics, the statistical values of ANOVA and its non-parametric double Kruskal-Wallis test have been compared. The normalization tests, i.e. the transformation values due to the improvement of distribution normality have been partially emphasized. There is a theorem in inferential statistics which contributes to making conclusions through the application of parametric statistics, even if normal distribution does not occur, i.e. the central border theorem. Its role in the study of education and upbringing has been specifically described.
Keywords
data distribution; normal distribution; kurtosis; skewness; normality evaluation tests; transformations; central border
Hrčak ID:
82182
URI
Publication date:
30.6.2011.
Visits: 12.643 *