24小时热门版块排行榜    

查看: 565  |  回复: 0

yeluping

至尊木虫 (著名写手)

[求助] 半变异函数

各位老师,想请教一个关于半变异函数的问题:
半变异函数的nugget/sill一般划分为:<25%强的空间相关性,25-50% 中等强,50-75% 中等弱,>75%弱。
我的半变异函数包括spherical,exponential,Gaussian三种模型,就放在一起根据nugget/sill的大小判断其空间相关性的强弱,结果老师指出:由spherical,exponential,Gaussian模型得到的nugget/sill是不同的,不能放在一起比较(不知道这样理解对不对,所以将老师的原话贴在下面),请问这个是什么意思啊?因为我看文献里面没有将不同模型的nugget/sill分开比较的。非常感谢了。

老师的原话:
The nugget-sill ratio needs to be handled with more care. You are using different types of models. Only for the spherical model, you have a clear transition between structured variability and the plateau. However, in the exponential and Gaussian models, the range is not exactly located at such a transition because in these models there is none. Aslo, the C0 + C (sill) is only a parameter but does not reflect an exact sill. Beyond the range parameter in these exponential-type models, the semivariance still increases further. Therefore, a nugget-sill ratio derived from spherical model parameters is different than one derived from an exponential or a Gaussian model!
回复此楼
已阅   回复此楼   关注TA 给TA发消息 送TA红花 TA的回帖
相关版块跳转 我要订阅楼主 yeluping 的主题更新
信息提示
请填处理意见