How to deal with event/estimation windows that overlap previous/future events?

[Jearlon Martina]

For my research I'm analysing a panel data set of monthly actual share repurchases.
The goal is to compute CARs for both monthly stock and bond returns around these repurchases.

Estimation window = (-24 months, -3months)
Event windows: (-3mo,+3mo), (-2mo,+2mo), (-1mo,+1mo)

An example of the data I'm working with:

Date	Company	Event
Month 1	A	1
Month 2	A	0
Month 3	A	0
Month 4	A	0
Month 5	A	1
Month 6	A	1

In Month 5 I see two things happening:

• The estimation window for the event in Month 5 will include the event in Month 1.
• The event windows for the event in Month 5 will include the event in Month 6

How do I deal with these kinds of overlaps in STATA?

• Should I just eliminate observations that have overlapping event/estimation windows? (How would I go about coding this?)
• Should I continue my analysis on the full sample and then another with those observations eliminated to see if results are significantly different?

[Jearlon Martina]

In the case that elimination of overlapping observations is the way forward, I can think of one solution for
the estimation/event windows and that is to replace event==1 in previous observations (up to -24 instead of -12) like shown in the picture above
However, I feel there is a much easier (read; more efficient) way to achieve my intended result