零误差实验

Zero-error Experiment

零误差实验可以更有效地观察粗差在经过抗差估计以后的分布情况,再将估计值与真值比较,从而较客观地对各种抗差估计方法做出评价。粗差相对于正常误差而言,正常误差可以不予考虑。本文通过一个三角网实验结果表明:方向平差和相关角度平差,对于经典平差(即最小二乘平差法),两种方法的结果一致,对于抗差估计,两种方法的结果不一致,但都是有效的;正态密度平差法优于Huber法、丹麦法、IGGI法、IGG法和Hampel法等抗差估计方法。

Zero -error experiment can observe the distribution of gross error after robust estimation, then compare the estimation value and real value, so we can evaluate externally all kinds of robust method. Natural error can not be considered relative to gross error. In this paper, through a trigonally net experiment we conclude： The values of directional adjust and correlative angle adjust is coherent in the least square method, it is different in the robust method, but they are effective. The normal density adjust excel the humber method, Danmark method, IGGI method, IGGIII method and Hampel method, and so on