Abstract:
The objective of this research was to compare on the power of some goodness of
fit test statistics for Log-normal distribution and Weibull distribution. The test statistics
are the Anderson Darling Test Statistic (AD), the Kuiper Test Statistic (K), the Ration of
Maximum Likelihoods Test Statistic (RML) and the Kolmogorov-Smirnov Test Statistic (KS).
The data for this research were generated from the two parameter Lognormal Distribution
and two parameter Weibull Distribution and the experiment was repeated 1000 times.
All parameter combinations were done with 20, 30, 40 and 50. Their efficiencies were
compared by controlling the probability of type I error and power of some goodness of fit
tests. The results of this study show that the Ratio of Maximum Likelihoods Test Statistic
(RML) best controls the probability of type I error best fit for the largest number data sets
using the Cochran criterion and the Kolmogorov -Smirnov Test Statistic (KS) best controls
the probability of type I error for the largest number data sets using the Bradley criterion.
The conclusion of research comparing the power of some goodness of fit test is that Ratio of
Maximum Likelihoods Test Statistic (RML) has highest power in all situations