Abstract:

The problem of reliability evaluation is usually handled based on the sample from the life testing experiment. If all the units in the life test are terminated at different times, it is called the nonequal type-I censoring. According to the nonequal type-I censoring sample, the existing methods usually estimate the parameter without the confidence interval. To fill this gap, a simpler method is proposed to construct the confidence limit on the basis of the order relations in the sample spaces under the exponential distribution. As comparisons, two approximate methods are also proposed based on the bootstrap method and the property of maximum likelihood estimate. Then, a Monte Carlo simulation and an illustrative example are conducted to compare the results of the three methods, which demonstrate that the new method outperforms the others.