< >
If you examine the reshaped data created by this logic, I think you
will find that only the last observation (with tradedatenum == 1382)
has a 1 in eventindicator. It is thus a 'singleton dummy'. Putting in
a dummy that is 1 for only 1 observation in a timeseries in plain old
OLS is equivalent to dropping that observation from the regression, as
you can then 'explain' y[1382] perfectly -- it has its own intercept
term. I don't think you want to do this.
This is not the standard methodology for an event study (in
particular, because the use of -sureg- prevents you from examining any
number of firms). Why wouldn't you rather want to estimate over the
pre-event period and forecast returns over the event period, and look
for abnormal returns?
Kit Baum, Boston College Economics and DIW Berlin
http://ideas.repec.org/e/pba1.html
An Introduction to Modern Econometrics Using Stata:
http://www.stata-press.com/books/imeus.html
On Nov 4, 2008, at 02:33 ,Thomas wrote:
scalar EventBegin = 1381
scalar EventEnd = 1382
scalar ObservationPeriod = 250
keep if TradeDateNum >= EventBegin - ObservationPeriod
keep if TradeDateNum <= EventEnd
gen EventIndicator = 0
replace EventIndicator = 1 if TradeDateNum >= EventBegin
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
