_______________________________________________Am 08.10.2024 um 12:38 schrieb Theodoros Panagiotidis:
OK - I find the following links more useful, however:
- https://dimewiki.worldbank.org/Event_Study
- https://theeffectbook.net/ch-EventStudies.html
Basically, to me it's a catch-all phrase. All depends on the time series model you find appropriate for predicting the non-event path (constructing the counterfactual), in order to compare this forecast with the actual realizations. And for inference you need to specify the functional form that the event effect reasonably takes. This could just be a step (=shift) dummy, or a broken trend, or an interaction, or whatever. Those are decisions that the researcher needs to make.
OK, I see that for teaching purposes a nice concrete example dataset with a known event, making whatever other technical assumptions, would be useful. If you (or somebody else) points us to such a dataset in whatever format, I'm sure that the needed few lines of code wouldn't take long to appear.
cheers
sven
Gretl-users mailing list -- gretl-users@gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave@gretlml.univpm.it
Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/