Testing exponentiality based on Type-I censored data

Abstract

The problem of goodness-of-fit for the exponential distribution when the available data are subject to Type-I censoring is discussed here. A test procedure is proposed in this case that exhibits more power as compared to existing methods. The power of the proposed test is assessed for several alternative distributions by means of Monte Carlo simulations. Finally, the proposed test is illustrated with a real data set.

Keywords:

• Anderson–Darling statistic
• Cramér–von Mises statistic
• empirical power
• exponential distribution
• goodness-of-fit testing
• Kolmogorov–Smirnov statistic
• Monte Carlo simulation
• mixture distribution
• order statistics
• Type-I censoring
• uniform distribution

Acknowledgements

The authors thank the anonymous reviewer and the Editor-in-Chief, Professor Richard G. Krutchkoff, for their useful comments and constructive suggestions on an earlier draft that led to the improved version of this paper.