【摘要】 该文受Berkson将检验方法用于估计未知参数的启发,根据三个拟合优度统计量导出三种新的求取等值系数的方法,即:平方根等值方法(Square Root criterion,SQRTcrit)、对称相对熵等值方法(Symmetric Relative Entropy criterion,SREcrit)、加权等值方法(Weighted criterion,Wcrit),即Haebara准则的加权式。虽然在被检验的两个分布列很接近时,这三个多项拟合优度检验方法是渐近等价的,然而用它们求取等值系数时,Monte-Carlo模拟结果表明这三种新等值方法的行为表现存在差异。它们之间的差异和随机误差的大小有密切关系,即与项目参数估计的精度有关;还与等值系数A的范围有关。更多还原

【Abstract】 This paper is inspired by applying Test Statistic to estimate unknown parameter,three new solving the equating coefficients methods(for short,equating method) are proposed in the light of goodness_of_fit test statistic,that are Square Root method,Symmetric Relative method,Weighted method which is the Weighted Haebara method.When two distributions are approximate,the three goodness_of_fit test statistic are near equivalent.But what is the result about the three equating methods derived from test statistic? Monte-Carlo study shows that there are differences among three equating methods.The difference has closely relationship with estimation random error and the domain of equating coefficient A.更多还原