Abstract:

Objective Multicollinearity among common temperature parameters was analyzed to get suggestions for establishing the temperature driving model of insect population. Methods SPSS 13.0 was used to analyze pairwise correlation of daily average temperature (X₁),maximum temperature (X₂), lowest temperature (X₃), 18:00 temperature (X₄) and five days' average temperature (X₅), which was collected from 2010 March to 2011 November. Meanwhile multicollinearity of temperature parameters was diagnosed based on mosquito population model droved by temperature. Results There was an extremely significant (P < 0.01) linear correlation between the various parameters. The coefficient of determination between X1 and X2 is the highest (R_c²=0.959); and it was the lowest between X4 and X5 (R_c²=0.811). Conclusion There were extremely significant correlation between any two temperature parameters discussed in this paper. These five temperature parameters had strong multicollinearity in mosquito temperature driving model. Multicollinearity weakened the model regression goodness. Simultaneous application of multiple temperature parameters should be avoided when introducing the temperature factor to establish temperature driving model.

Key words: Air temperature, Collinearity, Model, Mosquito

摘要：

目的 了解常用气温参数的多重共线性,为应用相关参数建立虫量的温度驱动消长模型提供参考。方法 应用SPSS 13.0软件分析无锡市2010年3月至2011年11月每天的平均气温(X₁)、最高气温(X₂)、最低气温(X₃)、18:00时气温(X₄)、5日均温(X₅)5个参数之间的相关性,并以本地蚊虫为例经向后法诊断了气温参数在蚊密度消长模型中的多重共线性。结果 各参数之间呈直线相关,相关性均有统计学意义(P < 0.01)。其中X₁与X₂的相关性最高,校正决定系数(R_c²)=0.959;X₄ 与X₅相关性最低,R_c²=0.811。在以气温参数为自变量建立蚊密度消长模型时,随着自变量个数的增加,多重共线性程度也增强,模型的复相关系数(R)升高,R_c²降低。结论 气温参数之间存在显著的自相关,在建模时存在较强的多重共线性,多重共线性实际上降低了回归方程的拟合度,在引入气温因子建立相关模型时应避免同时应用多个气温参数。

关键词: 气温, 共线性, 模型, 蚊虫