摘要： 采用聚类代表无法筛选自变量，根据逐步判别分析和逐步回归原理。通过计算机模拟技术，对河北省麦蚜主发生区田间系统调查资料进行了统计分析，组建了麦蚜发生程度4级判别式。应用该式可提前2个月预测穗期蚜量。经对保定、正定的历史数据回判。准确率达100%。对麦蚜种群数量增长研究结果表明，5月份百茎蚜量随时间变化呈“S”型。可用Lgistic曲线方程表述。本研究结果为麦蚜综合防治提供了决策依据。

关键词: 麦蚜, 聚类代表元, 统计分析, 计算机模拟, 预测模型

Abstract: The field census data of aphid population in Zhengding county and Baoding area, Hfibei province, was statistically analysed by use of the principle of stepwise regression and stepwise discrimination and the method of representative elements of clustering analysis for selecting the independent variables, and the four-grade discrimination formula for reporting the aphid disaster degree and the long- term prediction models for reporting the highest aphid density in May were established.The aphid population at the date of heading could be predicted two months ahead of time.In order to determine the reliable control date, the growth regularity of aphid population in May was studied.The results showed that the aphid population growth curve assumed a "S" type, and the logistic curve equation was Y =7764.519/1+2843.669e-1.3202x.The above results can provide basis for controlling aphid in winter wheat field.

Key words: Wheat aphid, Representative elements for clustering analysis, Statistical analysis, Computer simulation, Prediction model