Rolling regressions

#1
Hi everybody,

I was wondering if any of you have any experience with rolling regressions and how to calculate significance when using them.

I am trying to check whether one portfolio of stocks is better than another, corrected with some common factors of stock returns. To check this I want to do an OLS regression with a rolling 48 month window over a long period of time. After doing this in Stata, I end up with a time series containing the coefficients from all the iterations. The question is: How do I estimate the significance of the regression coefficient, i.e. if one portfolio outperforms the other. Can I simply run a normal t-test on the new coefficient time-series to check whether the average coefficient is significantly greater than zero? Or do I have to use a different test statistic altogether?

Any help you can offer will be greatly appreciated! If you feel the question needs clarification, I will gladly provide more information.