Exchange of censorship types and its impact on the estimation of parameters of a Weibull Regression model

The objective of this paper is to examine whether treating failure times, actually derived under type-I censorship, but treated as though type-II censorship were used, leads to important reduction in efficiency of the maximum likelihood (ML) estimators of the parameters of a Weibull Regression model, frequently used in accelerated life testing studies. This paper shows that exchange of censorship types results in minor reduction in the efficiency of various estimators. Such a conclusion suggests that results obtained under type-II censorship apply to corresponding (in terms of number of observed failures) type-I situations. The only exception to this general rule occurs when the number of test items at each level of stress drops below 5. >