为什么对non-linear regression, R square 没有意义?
所有跟贴
·
加跟贴
·
新语丝读书论坛
送交者: 老中乙号 于 2007-11-22, 11:45:30:
那该用什么parameter来表示regression的质量?
所有跟贴:
what journal/area have more published studies using non-linear reg?
-
blackbox
(88 bytes)
2007-11-22, 15:07:46
(185712)
phase I trials may need nonlinear models (无内容)
-
xj
(0 bytes)
2007-11-22, 16:17:12
(185724)
it depends what you mean R^2
-
xj
(660 bytes)
2007-11-22, 13:03:02
(185692)
R^2=1-SSreg/TSS. cannot get this correct. (无内容)
-
xj
(0 bytes)
2007-11-22, 16:12:51
(185722)
wrong (无内容)
-
xj
(0 bytes)
2007-11-23, 00:00:34
(185811)
did you ever publish your results from non-linear regression?
-
blackbox
(90 bytes)
2007-11-22, 14:58:27
(185711)
? logistic regression is nonlinear too
-
xj
(352 bytes)
2007-11-22, 16:11:44
(185721)
though both classical reg and logistic reg are members of
-
blackbox
(198 bytes)
2007-11-22, 21:57:50
(185783)
Compute R sqaured after linear transformation
-
HunHunSheng
(247 bytes)
2007-11-22, 12:07:47
(185663)
y=x^2 is still linear in the sense of beta, not the curve
-
xj
(29 bytes)
2007-11-22, 13:24:06
(185698)
originalGaussianerror after transform maynot thus ms maynotmake sense (无内容)
-
短江学者
(0 bytes)
2007-11-22, 12:23:11
(185671)
反正本来实际例子就是靠近似.理想的高斯分布本来不存在
-
HunHunSheng
(42 bytes)
2007-11-22, 12:37:15
(185682)
the goodness of fitting? (无内容)
-
PoohHunny2
(0 bytes)
2007-11-22, 12:00:14
(185658)
只是不好算吧?意义还是一样的。当然追不道的N有人就说不美 (无内容)
-
短江学者
(0 bytes)
2007-11-22, 11:51:14
(185654)
加跟贴
笔名:
密码:
注册笔名请按这里
标题:
内容: (
BBCode使用说明
)