中文摘要：

等价权抗差估计法保留最小二乘估计处理正常观测值的优良特性,但其抗差性与初值关系极大,若用最小二乘估值作初值,必定会影响其抗差性。中位数法具有很好的抗差特性,但它只用部分观测数据计算参数估值,丢失大量有效信息。基于中位参数的抗差估计方法,在有限样本时,给出其崩溃污染率的估算方法。根据中位参数法和等价权抗差估计法的各自优点,先用中位参数法计算初值,再用等价权抗差估计法计算最终的参数估值。试验结果表明：该方法比基于最小二乘初值的等价权法具有更强的抗差性,比中位参数法的计算精度高。

英文摘要：

The robust estimation based on equivalent weights keeps the outstanding properties of least squares（LS） adjustment in processing normal observations.Nevertheless,its robustness is highly related to the initial values.If the LS estimates,which are insensitive to outliers,are used as initial values it will certainly impact the robust characteristics of equivalent weight robust estimates.Comparatively,the least median（LM） estimates are much robust than LS estimates,but they are computed only with parts of observations,and many available observations are not used.We put forwards the median parameter robust estimation and present an approximate method to estimate its breakdown point in finite observation cases.By integrating the advantages of median parameter estimation with equivalent weight robust estimation,we take the median parameter estimates as the initial values and carry out equivalent weight robust estimation to compute the final results.The numerical experiments show that our robust estimation method of combined median parameter and equivalent weight is much better than individual robust estimations or median parameter estimation.