This thesis is concerned with the problem of designing step-stress accelerated life tests(ALTs) plans for Weibull distribution under Type I censoring. It is assumed that not only Weibull scale parameter but also Weibull shape parameter is a log linear function of (possibly transformed) stress, and that a cumulative exposure model holds for the effect of changing stress. Minimizing the asymptotic variance of maximum likelihood estimator of a stated percentile at design stress is used as an optimality criterion. When the number of stress level is given, optimum test plans - stress levels and stress change times - are obtained and the optimum three step plans are presented for selected value of design parameters. These plans are compared with the corresponding constant stress ALT plans in terms of relative efficiency. When the ratio of shape parameters at highest and design stress levels is known, the optimum simple step-stress ALTs with nonconstant shape parameter are also compared with those with constant shape parameter. Finally, the effects of errors in preestimate of the design parameter are investigated.
