Exercise: Treatments/outcome as trajectories
- Both treatment and outcome are measured at t = 32 (gray polygon). Outcome scale is omitted.
- Q: What would the data look like? How many rows/variables?
- Q: What conclusions would we draw from those measurements on the effect of university (0/1) on income (0-3000)?
- Q: Why is that conclusion problematic?
- Q: What considerations does the graph trigger if we extrapolate this problem to a larger population?
Lessons
- Lessons to be learned
- Ideally, we know the timing of when units moved from control to treatment as well as when the outcome changed.