Comparison of Power of Some General Goodness-of-Fit Tests

The comparison of a power of general goodness-of-fit tests of the Pearson's, the Kolmogorov-Smirnov's and of the author of the present paper has been presented.It is shown that for distributions like simular to exponential, the goodness-of-fit test of the author is locally the most powerful. The power of the given criteria for other alternative distributions has been investigated. It is shown that the Kolmogorov- Smirnov criterion in all considered cases is much poorer than the other two general criteria. It has been suggested that only criterion that gives the least value of a significance level of goodness-of-fit tests should be used in practice