I am using the Fama Macbeth approach for estimating coefficients. The two packages I've found on Stata (fm and xtfmb) don't provide root mean square error or adjusted R-square, perhaps because the way it is calculated does not lend itself to understanding how many degrees of freedom there are.
To simplify what I've tried so far (and for those who aren't familiar with FM):
-let's say my starting sample is spread across 7 years, with 100 observations each year (700 total). Say I have 5 variables. If I were to do a normal OLS, I'd do one regression with 700 observations and 694 degrees of freedom (700-5-1). Instead, though, using FM, I run seven different regressions (one for each time period). This yields 7 different values for each coefficient. I then take the mean of these seven values. The real FM goes on from there to determine if these values are significant from 0, but I take these 5 mean coefficients to estimate my dependent variable. My question is how many degrees of freedom I have.?. Or perhaps somebody understands the fm or xtfmb packages to help me find how to implement it?
(and I'm using Stata 13.1)
Thanks!
To simplify what I've tried so far (and for those who aren't familiar with FM):
-let's say my starting sample is spread across 7 years, with 100 observations each year (700 total). Say I have 5 variables. If I were to do a normal OLS, I'd do one regression with 700 observations and 694 degrees of freedom (700-5-1). Instead, though, using FM, I run seven different regressions (one for each time period). This yields 7 different values for each coefficient. I then take the mean of these seven values. The real FM goes on from there to determine if these values are significant from 0, but I take these 5 mean coefficients to estimate my dependent variable. My question is how many degrees of freedom I have.?. Or perhaps somebody understands the fm or xtfmb packages to help me find how to implement it?
(and I'm using Stata 13.1)
Thanks!
