This thesis is concerned with the problem of estimating the model parameters and percentiles of the lifetime distribution by the method of maximum likelihood with data obtained under accelerated lifetests in which two stresses $X_1$ and $X_2$ are constantly applied to the test items. It is assumed that ⅰ) lives of items follow a Weibull distribution, ⅱ) the Weibull log scale parameter is a multiple linear function of either $X_1$ and $X_2$ or $X_1, X_2$ and $X_1X_2$, and ⅲ) the Weibull shape parameter is independent of stresses.
Items are randomly assigned to each of the four combinations of stress variables ($X_1$ and $X_2$) and stress levels (low and high) according to an allocation rule, and tested until censoring time. For each of 120 different combinations of ALT setting of three allocation rules (1:1:1:1, 2:3:3:4 and 1:3:3:5), two low stress levels (0.1 and 0.3), four censoring times (0.4, 0.8, 1.2 and ∞) and five sample sizes (12, 24, 36, 48 and 60), lifetime data are generated and properties of the maximum likelihood estimators of the parameters and the percentiles at use condition are investigated with respect to biases and mean square errors by a Monte Carlo simulation.
