基于Partial-EIV的数据探测法研究

Data Snooping Research Based on Partial-EIV Model

为了解决EIV(Errors-In-Variables)模型的粗差问题,在Partial-EIV模型的基础上拓展数据探测法,对Partial-EIV的模型在(a_(0),β_(0))处进行线性展开,变换为最小二乘形式,进一步推导EIV模型下的单位权方差估计值,推导出数据探测法,通过算例分析,在观测值独立等精度情形下,可以有效地发现观测数据中的粗差。

For the problem of outliers in EIV(Errors-In-Variables)models,the data snooping method is extended based on the partial-EIV model.The Partial-EIV model is linearly expanded at(a_(0),β_(0)),and rewritten in the form of Least Squares,then unit weight variance estimation under EIV model is further deduced and data snooping method is derived.Numerical examples reveal that the method presented in this paper can effectively find the outliers in the observed data when the observed values are independent and with equal precision.