Code for out of sample R-squared?

[qi zhang nnn]

Hi all. I'm writing a thesis and I need to calculate out-of-sample R-squared, taking the form:

r_t is regressand; r_t_bar is average of the reggresaand up to time t; r_t_hat is the predicted value of r based on regression using data from t=1~t-1

I'll need an initial sample to for predicting r_hats. So basically, after using data on the first, say, 180 observations, I predict a r_181_hat, this is the r₁_hat in the formula; then for each of the next observations, I regress r on regressors using all available data up to t, predict the next period's value, etc etc. So may I have some suggestion how to code in Stata to recursively do the regression using Stata and generate one-period ahead r_hat? Thank you.

---

[Clyde Schechter]

There is an official Stata command, -rolling- which will do recursive rolling regressions. There is also Robert Picard, Nick Cox, and Roberto Ferrer's -rangestat-, available from SSC. Personally, I prefer the latter. The help file has a worked example.