讨论了利用历史数据估计线性回归模型ΔＷ＝ＳＣ＋η中参数Ｃ的验前分布Ｎ（Ｃ０，Ｋ０）的均值Ｃ０与方差Ｋ０的问题，得出了当Ｃ具有验前分布Ｎ（Ｃ０，Ｋ０），即均值为Ｃ０，协方差矩阵为Ｋ０的正态分布时，Ｃ的验后概率密度ＰＣ｜ΔＷ，改进了文献［２］中的ＢａｙｅｓＡ－值估计式，并对ＰＢ估计、广义岭估计、主成份估计、改进的主成份估计及广义压缩估计等５种有偏估值精度提出了评价方法。 This paper discusses the use of historical data to estimate the mean C0 and variance K0 of the pre-prioritized distribution N (C0, K0) of the parameter C in the linear regression model ΔW = SC + η. It is found that when C has the pretest distribution N (C0, K0) That is, when the mean value is C0 and the covariance matrix is ​​the normal distribution of K0, the post-test probability density of C is PC | ΔW, and the BayesA-value estimation formula in [2] is improved. The PB estimation, generalized ridge estimation, Component estimation, improved principal component estimation and generalized compression estimation.