Regression Discontinuity: Looking at People on the Edge: Causal Inference Bootcamp

Here we introduce the basic idea behind regression discontinuity. We also discuss the first empirical paper which both invented and used this technique.
Part of Duke University's Causal Inference Bootcamp: watch modules on this topic and much more at modu.ssri.duke.edu

[@Patrieees1]

Thank you so much!!!! You're soooooo much better than my economics prof.............

[@larazaremba1372]

Best explanation ever! Thank you so much!! :)

[@amyz.1033]

Very helpful! Thank you so much!

[@juancarlossquitieri5402]

thank you for this video! Finally I got to undertand RD!

[@laurendivenanzo734]

Very helpful and great explanation! Thanks!

[@DukeEHDi]

Thanks!

[@ericyaoaglozo9197]

This is very helpful. Thanks!

[@netsurfer5436]

Thank you!

[@DukeEHDi]

You're welcome!

[@yolamack7444]

Thank you so much!!!

[@DukeEHDi]

We're glad it helped you!

[@ghenyh2013]

thank you so much!!!!

[@emilyf7525]

I really need bootcamp like this for every subject lol.

[@adeelbutt2091]

thank you

[@DukeEHDi]

Good luck!

[@Levay92]

Thanks dude

[@WilliamPeck1958]

good job! really excellent, I'm just learning this aspect. *Skeptic questions I have: a) how was the data collected? I'd love to see the dataset. b) before the study, you could have predicted the outcome because it aligns with reality

[@Ag8MrE]

That was clear, and easy to follow. But if test-score is a confounder, why not just run an ANCOVA between the two groups, controlling for test score?

[@KrystalSarcone]

I should go read this original paper, but I had three questions: 1) Had the difference between the two point estimates at threshold been negative, it would have still implied a treatment effect, just not the one (or in the positive direction) hypothesized, right? 2) Why were national merit scholarships not considered in this? Did they not receive certificates of merit? 3) Has someone included national merit-based scholarship data and assessed if this RD treatment effect is statistically significant?

[@zouhierk933]

Yes, you are right, you have to read the original paper.

[@gugupiano2666]

So given those wiggly lines, if the results are not statistically significant, it is likely that there is actually no treatment effect in this paper?

[@thonghuynh7962]

Sir, I don't know who you are, but I'm gonna find you and thank you

[@irinucam]

What I don't understand is why can't they just take the same types of people and put one into a control group and one into the experimental one?