摘要
Logistic判别分析（LDA）可以认为是Logistic回归的预报形式，LDA还可由广义线性模型（GLIM）大大简化，对所有观测数据，GLIM可用矩阵形式描述，Logstic判别的系数可用Newton－Scoring算法得到，用ML-估计对LDA外生变量的逐步筛选是在对β作似然比检验的基础上进行的（称为逐步Logistic判别分析SLDA）。我们用数值预报产品建立了中国南京3－5月逐日的逐步LOGISTIC动力统计释用天气预报模型。
The Logistic Discrimination Analysis (LDA) can be regarded as the forecasting form of the logistic regression. The LDA is greatly facilitated by the well-known class of generalized linear models (GLIM). For all observations the GLIM can be described by a matrix form.The logistic discrimination coefficients can be obtained by Newton-scoring algo-rithm. A Stepwise selection of exogenous variables for LDA (SLDA) using ML estimation is based on likelihood ratio tests for the β. We use the SLDA to model the Dynamic-Statistic Interpretation Forecast Model for weather situation from March to May in Nanjing, China using NWP Products
出处
《气象科学》
CSCD
北大核心
1996年第1期81-85,共5页
Journal of the Meteorological Sciences
基金
"八五"国家科技攻关项目!85-906-05-02
