海洋水团的Fisher型逐步判别预报

判别分析是多元分析方法之一,我们将其应用于海洋水团的判别预报。用逐步判别的方法挑选因子,依每个因子的Wilks量进行F—检验,逐步地将判别“能力”强的因子选入。同时将由于其他因子组合而相形见绌者及时剔除。根据矩阵变换的信息,给出了Fisher型逐步判别预报式。将北黄海冷水团强度预报作为示例,给出了筛选因子及建立判别式的全过程。用历史资料检验了预报效果。与习用的Bayes型逐步判别预报进行了对比。

A SUCCESSIVE DISCRIMINATION FORECAST OF MARINE WATER MASSES

Discrimination analysis is one of the multivariate analysis methods. We apply it to the discrimination forecast of marine water masses. A method of successive discimination is utilized to select factors. The F-test is made with reference to .the Wilks value for each factor. The factor which has great "ability" of discrimination is selected successively. And a factor is immediately disqualified from discriminant when its "ability" declines because of the combination of given factors. A Fisher successive discriminant of forecast is given with reference to the information of matrix transformation. A multistage strength forecast of the Northern Huanghai Sea Cold Water Mass is taken as an example. The whole process of sifting factors and establishing discriminant is given. The effect of forecast is tested by historical data. A comparison is made between the above method and the conventionally used Bayes successive discrimination forecast.