This article presents a general form of the estimator for identifying dispersion effects from unreplicated two-level factorial experiments,and shows that the widely used estimators such as the BH,MH,and AMH estimators...This article presents a general form of the estimator for identifying dispersion effects from unreplicated two-level factorial experiments,and shows that the widely used estimators such as the BH,MH,and AMH estimators are all special cases of the proposed one,designated as the G estimator.The unbiased condition of the G estimator is proved,and a lower bound of variance of the G estimator is provided.A simulation based on a realistic design illustrates the variation of the variance and MSE(mean square error) of the G estimator on different coefficients.This estimator may be more flexible and has better performance than other methods such as the BH and MH estimators by appropriately selecting the coefficients.展开更多
基金This work is partially supported by the NSFC(No.10471157)on the part of the second author is also supported in part by the Advanced Research Center of the Sun Yat-Sen University(No.06M11).
基金2013 Major National Development of Special Equipment(No.2013YQ160487)2013 Beijing University of Post and Communication 863 Program(Nos.2013AA013403,2012AA010407)the Natural Science Foundation fo China(No.61007043)
基金supported by the National Natural Science Fund of China under Grant No.61503228Scientific and Technological Innovation Programs of Higher Education Institutions in Shanxi under Grant No.2015106
文摘This article presents a general form of the estimator for identifying dispersion effects from unreplicated two-level factorial experiments,and shows that the widely used estimators such as the BH,MH,and AMH estimators are all special cases of the proposed one,designated as the G estimator.The unbiased condition of the G estimator is proved,and a lower bound of variance of the G estimator is provided.A simulation based on a realistic design illustrates the variation of the variance and MSE(mean square error) of the G estimator on different coefficients.This estimator may be more flexible and has better performance than other methods such as the BH and MH estimators by appropriately selecting the coefficients.
