st: re: missing observations in xtabond/xtabond2

[Kit Baum <baum@bc.edu>]

In a posting yesterday I suggested that xtabond and xtabond2 routines  
might have problems with gaps in the timeseries within one or more  
panels. The implementations of DPD in Ox and GAUSS certainly have  
such problems (as the author of one of those routines has told me). I  
have been assured that official Stata's xtabond does not suffer from  
this flaw (it was revised some time ago to correctly handle internal  
gaps). I cannot say whether the user-written xtabond2 is sensitive to  
this issue. Essentially the challenge is to recognize that an  
'interrupted' time series for a single unit still corresponds to a  
single unit's data, and does not reflect two different units' data.  
Perhaps David Roodman can speak to this issue with respect to his  
xtabond2 implementation.

A more precise statement would be that the A-B (and B-B) estimators  
can be defined for interrupted time series, but not all  
implementations of A-B or B-B will deal appropriately with such a  
data set.

Kit

Kit Baum, Boston College Economics
http://ideas.repec.org/e/pba1.html


*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/