Power of tests of normality for detecting scale contaminated normal samples
Titel:
Power of tests of normality for detecting scale contaminated normal samples
Auteur:
Thode, Henry C. Smith, Laurel A. Finch, Stephen J.
Verschenen in:
Communications in statistics
Paginering:
Jaargang 12 (1983) nr. 6 pagina's 675-695
Jaar:
1983
Inhoud:
Twenty-seven tests of normality or goodness of fit were evaluated for power at detecting a scale contaminated normal sampleo This study used 1,000 replications each of samples of size 12, 17, 25, 33, 50, and 100 from six different contaminated normal distributions. Moment-type tests (e.g., kurtosis) were the most powerful over all sample sizes and contaminations. The Kolmogorov-Smirnov, chi-squared, Anderson-Darling, and Cramer-von-Mises tests had very low power at detecting contaminated normal random variables. Tables of the power of the tests and the power curves of certain tests are given.