What you will learn
- Pre-event and post-event windows
- Control observations
- Time-of-day effects
- Exclude rules defined before results
Pre-event and post-event windows
Define pre-event and post-event windows before inspecting outcomes. Different horizons answer different questions. Choosing the best-performing interval afterward is another form of model selection that must be disclosed.
Control observations
Control observations help describe ordinary behavior under comparable conditions. Matching by time of day or other predetermined features can reduce obvious differences, but it does not automatically establish causal identification.
Time-of-day effects
Time-of-day effects can influence volume, spread and volatility independently of the event. A comparison between event mornings and quiet nights can confuse session patterns with event-associated changes.
Exclude rules defined before results
Exclusion rules should address known data defects or overlapping conditions symmetrically. Removing only events with inconvenient responses produces a biased sample, even if each removal has a plausible retrospective story.
Worked example
An analyst examines one-, five-, fifteen- and sixty-minute responses and reports only the most favorable horizon as if it was chosen in advance. The horizon search is part of the experiment.
Try it yourself
Write a protocol that either preselects one primary horizon or explicitly treats multiple horizons as exploratory.
Show the worked solution
Name the primary horizon and secondary analyses before evaluation, or disclose all tested horizons and validate any selected conclusion on later events. Do not rewrite the study history around the best result.
Apply this to your course project
Submit a preregistered event-study protocol and a reproducible worked sample.
Keep the calculation inputs, assumptions and decisions with your work. Practical exercises are self-reviewed; the scored knowledge checks assess the questions shown, not an independent certification of practical competence.