一种模型模拟结果的统计检验方法

A statistical method for validation of ecological models

模拟结果的准确性是衡量生态学模型是否成功的关键,但采用统计学方法判别模型模拟结果与观察值相符程度的报道较少。根据两个直线回归方程能否合并为一个方程的统计学检验方法,提出了通过检验观察值与模拟值直线回归方程和1∶1直线方程截距与斜率是否相同,进而在统计显著水平上判断生态学模型模拟值与观察值一致性的统计学检验方法。数据检验表明,此方法可以较好解决判断生态学模型模拟结果准确性的问题。

Statistical validation has rarely been conducted to examine the consistency between observed data and the output of ecological models,although such validation is critical for determining the goodness-of-fit of such models. Based on the statistical principle of whether two regression models can be combined,a statistical method is proposed for validating ecological models. In the method,a linear regression is fit to observed（ x-axis） and modeled（ y-axis） data,then tested for significant differences in slope and intercept from the line y = x. A case study shows that the method could be used to validate ecological models.