山东省小麦病虫害分区预测模型的研究

对影响１９８２￣１９９１年山东省及各类型区小麦病虫发生量（预报量ｙ）的环境因素（预报因子ｘ１），在一元线性回归初选的基础上，经逐步回归、逐步判别、多元回归和多级判别分析进行精选，共建立全省和各类型区预测模型１８６个，从中筛选出最优预测模型３０个，于１９９３年和１９９４年应用，平均预测准确率达９０％以上。

STUDY ON FORECASTING MODEL OF DISTRICTS OF WHEAT DISEASES AND INSECT PESTS IN SHANDONG PROVINCE

in this paper, the authors not only sorted out the systematic monitoring data anwheat diseases and insect pests and factors affecting their occurrence, but also analysed theforcasting targets y(occurring amount) and different combinations of forecastingfacts xj(j = 1, 2, """, m) that affecting the occurrence of wheat diseases and insect pests byadopting some statistic methods such as stepwise regression, discrimination step by step,multivariate regression and multiplestage discrimination. They established 186 predictingmodels of wheat diseases and insect pests in different types areas or over all provincethrough calculation different values of FI and F2 of stepwise regression and discriminationstep by step, importing constantly main factors that posses fine interrelation, eliminationsecondary factors that posses poor interrelation, choosing precisely foreasting factors onthe basis of primary choose with single factor linear regression. They screened out 30 "the finest" predicting mathematical models according to the ratio of back-checking with historymaterial of variety models and stabilities status. The average predicting accuracy is 90%through applying in 1993 and 1994.