小波辅助的拟准检定法和补偿最小二乘估计解算污染半参数模型

Resolution of Polluted Semi-Parametric Model with Wavelet Assisted Quasi-Accurate Detection and Penalized Least Square Estimation

利用拟准检定法探测粗差时,解算的污染半参数模型的标准化残差受到模型系统结构的干扰,难以准确初选拟准观测值。通过小波变换滤波分析的模极大值点初选拟准观测值,在拟准观测真误差估值范数极小的条件下实现对粗差的定位定值,最后采用补偿最小二乘估计分离系统误差,得到较为准确的参数估值。

The standardized residuals of polluted semi-parametric models have been influenced by model system structures when making use of the method of quasi-accurate detection of gross errors,so it is difficult to choose the quasi-observations in the primary selection in this way.In this paper,the observations are selected with the modulus maxima under the wavelet transform filter analysis as primary quasi-observations.Then the location and value of gross errors are realized on condition of the minimum of the real errors related to quasi-accurate observations.At last,penalized least square estimation is used to separate the systematic errors to obtain parameter estimation accurately.