On Wed, 23 Sep 2009, Henrique wrote:
I'm estimating some OLS equations that don't have a constant
term,
so, the R-squared needs to be adjusted, ok? What do you suggest? I'd
realized that EViews automatically adjusts the R-squared when the constant
term is absent (but the User's Manual doesn't tell how).
In the absence of an intercept gretl nowadays uses the formula
recommended by NIST,
R^2 = 1 - SSR/(sum of the y^2)
Another possibility, which gretl used to report, is the square of
the correlation between y and yhat.
Allin Cottrell