like1 发表于 2010-6-9 23:23 
sas中除了reg和glm过程中计算决定系数，还有那个过程可以计算决定系数。
可不可以在 nlin过程中输出决定系数呢？
[em43]
The usual R^2 definition is only valid in a classical linear regression under assumption of variance being homoscedascity. That is why you normally see it in PROC REG and PROC GLM. It is bounded [0,1] and provides a simple and clear interpretation. In nonlinear regression, for example logit or probit model, the assumption is no longer hold. The usual R^2 definition COULD be meaningless.
BTY, there are many other statistics could be used as model evaluations.
扫码加好友，拉您进群
全部版块
我的主页
收藏
[em06]
