摘要
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。
First, the history of systematic error analysis in the crossover error analysis is reviewed.It is pointed out that two indispensable stages should be included in analyzing the systematic error, i.e., the selection of the systematic error model and the estimation of the coefficients in the selected model, however, only the latter is addressed in most of the literature while the former is （even completely ） ignored. Second, three kinds of methods, widely used in model selection, i. e., the hypothesis test, cross validation, and the information theory based methods, are commented and compared, and the superiority of the latter is emphasized. Third, the information-theory-based model selection method is introduced mainly focusing on the famous Akaike information criterion.The formula calculating this criterion is given ,the relationships with other criteria or other model selection methods are analyzed, model averaging and model selection uncertainty evaluation based on this criterion are also introduced.
出处
《海洋测绘》
CSCD
2015年第5期1-7,共7页
Hydrographic Surveying and Charting
基金
国家自然科学基金(41404001
41374018)
