Analyzing Location And Dispersion Effects From Replicated Factorial Experiments With Right-censored Data

In replicated factorial experiments, censored data may be easier or less costly to collect than completed data because of some reasons, such as time and cost. Hamada and Wu (1991) identified important location effects by imputing the censored data. And Bihua has considered identification of both location and dispersion effects from unreplicated factorial experiment with right-censored data. But the method is constrained in application for replicated experiments.. A method is presented for estimating the location and dispersion effects from these experiments. Firstly, we estimate the variance and the mean of each cell with maximum likelihood; secondly, we identify the important dispersion effects based on least squares analysis of the logarithm of within-replication variance; last, we identify the important location effects based on weighted least squares analysis of the mean of each cell. A simulation study also demonstrates its superiority over some existing methods. An experiment for the robust design of thermostat is used to illustrate the method.