Analysis Of Location And Dispersion Effects In Unreplicated Factorial Experiments With Censored Data

It's crucial how to identify and estimate both location and dispersion effects accurately in factorial experiment analysis. The method of identifying and estimating the factor location effect is the main content of reseatch in the traditional experiment design with complete data (the observation of experiment is known exactly). In the standard method the identification of dispersion effects typically requires replications with the fixed factor levels. Thus,the analysis of dispersion effects from unreplicated factorial experiments has become an important research topic in recent years. Brenneman and Nair(2001) proposed a method to estimate dispersion effects(MH) through concluding several used methods and study the performance of the method through simulation. However the analysis of both location and dispersion effects in unreplicated factorial experiments with censored data has not been studied widely. Hamada and Wu(1991) proposed a iterated method to estimate location effect under the condition of homogeneity of variances. It's unable to estimate the dispersion effects.In this paper, we consider a model of heterscedasticity, and present a new algorithm on the basis of Hamada and Wu(1991)'s iterated algorithm on estimation of location effects, combining Brenneman and Nair(2001)'s MH method on estimation of dispersion effects. This algorithm of model selection and simutanious estimation of location and dispersion effects analyze unreplicated factorial experiments with censored data. The simulation study demonstrates the feasibility of the algorithm, and in the meaning of criteria of MSB the estimation accuracy of location effects is superior to Hamada and Wu's. Finally we make a further analysis of the Hamada and Wu's example by applying this algorithm.