Experimenters often employ censoring schemes in life tests to shorten the test time or to reduce the test cost. However, censoring restricts the ability to observe failure times exactly. If the lifetime distribution is not normal, incomplete information contained in censored data often causes diﬃculties with employing the experimental design methods to conduct statistical inferences in engineering applications. The paper develops three new modiﬁed maximum likelihood predictors (MMLPs) to predict type II censored data for the Weibull distribution. Once the censored data are predicted, the predicted information can be merged with uncensored data as a pseudo-complete data set. The robust 2^k factorial design method  can be employed to identify the signiﬁcant factors from the experiment for the pseudo-complete data set. A Monte Carlo simulation study is conducted to evaluate the performance of the proposed method, and an example regarding engineering is presented for illustration.
International Journal of Innovative Computing, Information and Control 5(12)pt.A, pp.4415-4429