比较R中的回归模型 [英] Comparing regression models in R

问题描述

我需要测试自2002-2004年以来1999-2001年期间的数字是否显着改善.我尝试将数据汇总到这两个时段中，​​并比较线性回归模型(例如lm(Y〜A + B))的调整后R平方，但这并不能得出正确的结论.我认为公司之间的回归将更为相关，因为回归系数自然会因公司而异.

I need to test if figures from the 1999-2001 period significantly improved since 2002-2004 period. I tried pooling the data into these two periods and comparing adjusted R-squares of a linear regression model (e.g., lm(Y~A+B)) but that does not lead to the right conclusion. I suppose a by-company regression would be more relevant because regression coefficients will naturally differ from company to company.

如何在R中进行这样的公司固定回归?还是有另一种方法可以测试我的模型在两个时期内是否变得更合适"?谢谢

How can I do such a by-firm regression in R? Or is there another way of testing if my model has become a 'better fit' over the two periods? Thanks

数据看起来像这样(当然会有更多的公司):

Data looks something like this (way more companies of course):

Company Year    Y           A             B
11308   1999    -0,0208100  0,014718891 -0,006672241 
11308   2000    -0,0073200  0,01513105  -0,001765405 
11308   2001    -0,0242500  0,026331427 0,011924914 
11308   2002    0,0071770   0,033910057 -2,55861E-05 
11308   2003    -0,0161000  0,039996572 0,003413556 
11308   2004    -0,0283000  0,038958565 0,004018833 
11850   1999    -0,0001400  0,044492288 0,008268478 
11850   2000    -0,0023400  0,057337917 0,028973756 
11850   2001    -0,0113100  0,049981605 -0,002928416 
11850   2002    0,0055080   0,04095854  -0,015228795 
11850   2003    -0,0150000  0,089150637 0,042316779 
11850   2004    0,0065680   0,093468014 0,016125354

推荐答案

听起来像从r到z的转换在这里可能效果很好: http://vassarstats.net/rdiff.html

Sounds like an r-to-z transformation might work quite well here: http://vassarstats.net/rdiff.html

R中的此程序包可以为您做到: https://cran.r-project.org/web/packages/cocor/cocor.pdf

This package in R can do it for you: https://cran.r-project.org/web/packages/cocor/cocor.pdf

这篇关于比较R中的回归模型的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！