p******e
发帖数: 17163 | 1 Multiple Linear Regression - Regression Statistics
Multiple R 0.955577964077842
R-squared 0.913129245431153
Adjusted R-squared 0.90733786179323
F-TEST (value) 157.670308603251
F-TEST (DF numerator) 3
F-TEST (DF denominator) 45
p-value 0
Multiple Linear Regression - Residual Statistics
Residual Standard Deviation 56.9198764270578
Sum Squared Residuals 145794.254961219
Meta Analysis of Goldfeld-Quandt test for Heteroskedasticity
Description # significant tests % significant tests OK/NOK
1% type I error level 14 0.388888888888889 NOK
5% type I error level 18 0.5 NOK
10% type I error level 20 0.555555555555556 NOK
线到挺fit的？ |
d**********r
发帖数: 24123 | 2 Adjusted R-squared 0.9
P-Value = 0
This line fit is too good to be true |
k**********4
发帖数: 16092 | 3 what is the question?
【在 p******e 的大作中提到】
: Multiple Linear Regression - Regression Statistics
: Multiple R 0.955577964077842
: R-squared 0.913129245431153
: Adjusted R-squared 0.90733786179323
: F-TEST (value) 157.670308603251
: F-TEST (DF numerator) 3
: F-TEST (DF denominator) 45
: p-value 0
: Multiple Linear Regression - Residual Statistics
: Residual Standard Deviation 56.9198764270578
|
d**********r
发帖数: 24123 | 4 Does not matter, I will reject anything like this in NDA
【在 k**********4 的大作中提到】
: what is the question?
|
p******e
发帖数: 17163 | 5 is this model trustworthy?
【在 k**********4 的大作中提到】
: what is the question?
|
k**********4
发帖数: 16092 | 6 model fitting is pretty good, but it does not
mean it is trustworthy. To be more specific,
it has internal validity, whether it has
external validity or not depends on whether
choosing this model makes sense or not.
【在 p******e 的大作中提到】
: is this model trustworthy?
|
p******e
发帖数: 17163 | 7 is there a way to test external validity?
【在 k**********4 的大作中提到】
: model fitting is pretty good, but it does not
: mean it is trustworthy. To be more specific,
: it has internal validity, whether it has
: external validity or not depends on whether
: choosing this model makes sense or not.
|
i*********n
发帖数: 5465 | |
b*****p
发帖数: 9649 | 9 求解释！谢谢！
座机，完全是门外汉。
用layman's word,如何解释？
【在 i*********n 的大作中提到】
: 多半是endogenous的
|
i*********n
发帖数: 5465 | 10
http://en.wikipedia.org/wiki/Endogeneity_(economics)
最简单的。转弯之前大部分人要打转向灯。所以打转向灯和转弯是90%以上的
correlation (p<0.000001)，而且转向灯发生在转弯之前。我们能不能说转向灯导致了
车子转弯? 不能，因为这打转向灯是内生的。
所以我从来不相信统计学出来的东西，no identification, no causality。。。现实
生活中的问题，model fit 太高的一般都有问题
哈哈
【在 b*****p 的大作中提到】
: 求解释！谢谢！
: 座机，完全是门外汉。
: 用layman's word,如何解释？
|
|
|
k**********4
发帖数: 16092 | 11 No, best thing u can do is simulations.
【在 p******e 的大作中提到】
: is there a way to test external validity?
|
b*****p
发帖数: 9649 | 12 谢谢info大牛！
所以简单的说
A&&B =\=> A->B
because A->B == ~A || B
寄信人: deliver (自动发信系统)
标 题: 本站转帐通知单
发信站: BBS 未名空间站 (Fri Mar 22 12:54:18 2013)
来 源: mitbbs.com
bullpop，您好：
您转给 information,现金(伪币):10,收取手续费:0.1
同时附加了如下留言给 information.
thanks for sharing!
站务 |
k**********4
发帖数: 16092 | 13 but u can say turning signal is highly correlated
with turning, that is a valid conclusion.
【在 i*********n 的大作中提到】
:
: http://en.wikipedia.org/wiki/Endogeneity_(economics)
: 最简单的。转弯之前大部分人要打转向灯。所以打转向灯和转弯是90%以上的
: correlation (p<0.000001)，而且转向灯发生在转弯之前。我们能不能说转向灯导致了
: 车子转弯? 不能，因为这打转向灯是内生的。
: 所以我从来不相信统计学出来的东西，no identification, no causality。。。现实
: 生活中的问题，model fit 太高的一般都有问题
: 哈哈
|
i*********n
发帖数: 5465 | 14
Science cares about causality
drawing conclusion based on association is very dangerous
that's why we need to have experiments!
【在 k**********4 的大作中提到】
: but u can say turning signal is highly correlated
: with turning, that is a valid conclusion.
|
p******e
发帖数: 17163 | |
k**********4
发帖数: 16092 | 16 I agree. but still correlation is a necessary (although not
sufficient) condition for causality.So there is information
there.
【在 i*********n 的大作中提到】
:
: Science cares about causality
: drawing conclusion based on association is very dangerous
: that's why we need to have experiments!
|
b*****p
发帖数: 9649 | 17 you can wait and hope.... LOL
【在 p******e 的大作中提到】
: 楼上两位真的都是女的吗？
: 我汗颜啊
|
