A method of combining two independent test statistics to test the homogeneity of several Weibull (or extreme value) populations under censoring is considered. The procedures developed in this paper are based on Fisher’s method of combining two likelihood ratio statistics and two score statistics. The performance of these procedures are examined, through simulations, in terms of empirical size and power of the test statistics. We also develop two other statistics to test the equality of several Weibull scale parameters and shape parameters (or extreme value location and scale parameters) simultaneously. Simulation results show that the Fisher’s method of combining two independent statistics performs reasonably well under censoring.
Likelihood ratio statistic Score statistic Fisher’s method Extreme value distribution Censored data
AMS 2000 Subject Classification
62N05 62N01 62F03
This is a preview of subscription content, log in to check access.
McCool, J. I. 1979. Analysis of single classification experiments based on censored samples from the two-parameter Weibull distribution. J. Stat. Plan. Inference, 3, 39–68.MathSciNetCrossRefGoogle Scholar
Paul, S. R., and K. Thiagarajah. 1992. Hypothesis tests for the one-way layout of type II censored data from Weibull populations. J. Stat. Plan. Inference, 33, 367–380.CrossRefGoogle Scholar
Singh, N. 1986. A simple and asymptotically optimal test for the equality of normal populations. A pragmatic approach to one-way classification. J. Am. Stat. Assoc., 81, 703–704.MathSciNetCrossRefGoogle Scholar
Thiagarajah, K., 1992. Inference procedures in some lifetime models, Ph.D. thesis, University of Windsor, Canada.Google Scholar
Thiagarajah, K., and S. R. Paul. 1993. C(α) tests for the analysis of one-way layout of data having extreme value distribution. Statistician, 42, 85–100.CrossRefGoogle Scholar