异方差对生物量模型构建的影响

The Influence of Heteroscedasticity on the Establishment of Biomass Model

生物量建模中的误差项往往存在异方差性,影响模型的可信度。通过介绍异方差检验和消除的方法,结合兴安落叶松的实测数据,借助SPSS统计软件实现了建模过程中对异方差的检验和消除,并达到了很好的估计效果。结合实例分析,采用模型转换和加权函数的方法对模型的预估效果起到了很好的作用。建立树干的相容性生物量模型,模型决定系数R2为0.947,估计值的标准差SEE为17.368,平均估计误差MPE为2.637,平均相对偏差ME为3.957,平均相对偏差绝对值MAE为0.408,以及预估精度P为92.39%,各项检验指标均符合建模要求。

There are often heteroscedasticity in the error term during biomass modeling, which affects the reliability of the model. Through the introduction to the method of heteroscedastieity testing and elimination, the measured Larix gmelini data was used to in- spect and eliminate the heteroscedasticity in the modeling process with the aid of SPSS statistical software, and very good estimation re- suits were achieved. Combined with case analysis, adopting the method of model transformations and weighted function has played a good role in model prediction. The compatibility trunk biomass model was established, The model determination coefficient R2 is O. 947 the standard deviation of SEE is 17. 368, the average estimated error MPE is 2. 637, the average relative deviation of the ME is 3. 957, the average relative deviation absolute MAE is 0. 408, and the prediction precision is 92. 39% , various test indexes meet the modeling requirements.