Hello everybody,
I am analyzing weekly state lottery sales. The dataset can be thought
of as a panel of 200 games observed over a total of 350 weeks, but not
all games enter and exit at the same time. Instead, a number of games,
between 1 and 5, are launched every week between July of 2001 and
March of 2007. Games live for 31 weeks on average. The longest-running
game is observed for 118 weeks.
Essentially, if you arranged the sales variable as a table of 200
columns ordered from the oldest game to the youngest, and 350 rows
numbered from the first week of the oldest game to the last week of
the youngest, this table would have a wide band of non-missing
observations following roughly the main diagonal, but the corners at
either end of the second diagonal would be empty swaths. The same goes
for most of the covariates of interest.
Does this missingness pattern in itself recommend any particular
estimation technique? In particular, can I still xtset the data along
i(1-200) and t(1-350) for a panel regression? Will xtreg assume that
the missing observations are missing at random? I don't think it
should, but what are the alternatives?
Thank you,
Gabi
*
* 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/