Further Results for Goodness-of-Fit Statistics with Unknown Parameters.
STANFORD UNIV CALIF DEPT OF STATISTICS
Pagination or Media Count:
The report revises and extends the report Asymptotic results for goodness-of-fit statistics when parameters must be estimated Stephens, 1970 new theoretical and numerical results are given. In particular the weights are calculated when asymptotic distributions of W squared, U squared, and A statistics associated with Cramer-von Mises, Watson and Anderson-Darling are expressed as weighted sums of chi square sub 1 variables. These lead ultimately to the provision of percentage points for W squared, U squared, and A in tests for normality and exponentiality, when parameters must be estimated. The previous report gave only means and variances of asymptotic distributions some of these are repeated for completeness. Author
- Statistics and Probability