海洋测量交叉点误差分析(二):系统误差模型的选择

首先对海洋测量交叉点误差分析中系统误差分析进行了回顾,指出系统误差分析是一个模型选择和模型参数估计的双重问题,而传统方法中大多仅涉及了后者而忽略了前者。然后对常用的三种模型选择方法,即假设检验、交叉检核、基于信息论的方法,进行了分析和比较,指出后者更具优越性。最后对以Akaike信息量准则为核心的基于信息论的模型选择方法进行了介绍,给出了Akaike信息量准则的计算公式,比较了同其他模型选择方法的关系,并介绍了基于该准则的模型平均方法和模型选择不确定性评估方法。

Analysis of Crossover Errors in Marine Surveys,Part II:The Selection of Systematic Error Model

First,the history of systematic error analysis in the crossover error analysis is reviewed.It is pointed outthat two indispensable stages should be included in analyzing the systematic error,i.e.,the selection of thesystematic error model and the estimation of the coefficients in the selected model,however,only the latter isaddressed in most of the literature while the former is ( even completely) ignored.Second,three kinds ofmethods,widely used in model selection,i.e.,the hypothesis test,cross validation,and the information theorybased methods,are commented and compared,and the superiority of the latter is emphasized.Third,theinformation-theory-based model selection method is introduced mainly focusing on the famous Akaikeinformation criterion.The formula calculating this criterion is given,the relationships with other criteria or othermodel selection methods are analyzed,model averaging and model selection uncertainty evaluation based on thiscriterion are also introduced.