I have a $(n \times d)$ panel $y$ of $n$ different variables , and a $(n \times d)$ panel $x$ of their forecasts.

$d=$ time length of data

$n=$ cross section width/ no. of variables

I run a pooled (panel) regression in three ways:

$1.$ simple pooled regression which simply aggregates the data across time and variables:


$2.$ regression of cross sectional means (so mean of all columns of y and x at each point of time):

$ols(y.mean(axis=1), x.mean(axis=1)$

$3.$ pooled regression of cross-sectionally demeaned variables:

$ols( (y-y.mean(axis=1) , (x-x.mean(axis=1)) )$

Can I expect to see some sort of relation between betas/t-stats/r-square of these regressions?

e.g. Should it hold that $R_1^2$ lies between $R_2^2$ and $R_3^2$ etc?

How about the case when I enforce the intercept to be 0?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.