On Wed, 12 Mar 2008, scottd.orr(a)comcast.net wrote:
The manual seems to suggest that gretl's normal procedure for
handling of lags makes the Arrellano/Bond estimator
unnecessary--but it doesn't give any detail on that procedure.
Or am I missing something?
1) Gretl will generate lags correctly in a panel dataset, entering
a missing value at the start of each time-series.
2) This doesn't really have much to do with Arellano-Bond. A-B is
a method for producing consistent estimates when a panel model
includes lags of the dependent variable. Simply getting the lags
correctly calculated does not produce this effect.
Allin Cottrell