Need help?

[pvalenzuelac]

Hi Nikos!
Thank you for this amazing package. For the final project of one of my courses I was planning to do a RDD package. I was surprised that there was no good RDD package in Julia. Then, I found yours so I decided to change a little bit my plans and, instead of creating a second RD package, use my time to improve your RegressionDiscontinuity.jl package. After talking to the course's prof, he recommended me to ask you what are your short terms plans for the package and try to help on this list instead of just submitting a pull request (he argues that it will be easier for everyone). The expected length of the project should be similar to @evanmunro 's pull request, but I will be happy to continue contributing over time. If you don't have a to do list, then I will work on what I think is the next step (probably trying to adapt Calonico, Cattaneo, and Titiunik's optimal bandwidth) and submit a pull request.

[nignatiadis]

Hi Pablo,

It is great to see interest in this package and I agree, it would be better for the community to have just one package that implements all major methods for RDDs. We have been wanting to write a comprehensive TODO list with @erikcs, but I have delayed this. What is your timeline for the project and do you have any idea in what aspect of RDs you would be most interested in contributing?

Some relevant thoughts (and will try to provide more details later): 1) Regarding the CCT bandwidth; it seemed that @nicolas-jimenez also has written some code. 2) Another alternative could be to implement a manipulation test, e.g., the one by McCrary [1]. The advantage here would be that this contribution would be relatively independent from any (breaking) changes that might happen to the package as we merge the PR of @evanmunro. 3) This package is MIT licensed (and I would like to keep it this way). This means in particular that you cannot look into the code of GPL licensed packages to figure out how to implement a specific method (even packages in other languages), but you can of course compare that the output is the same/similar.

Best,
Nikos

[1] McCrary, J. (2008). Manipulation of the running variable in the regression discontinuity design: A density test. Journal of econometrics, 142(2), 698-714.
 

[pvalenzuelac]

Hi Nikos,
Thank you for your quick reply!
Regarding my timeline: the due date for my course project is December 18th. However, I would love to continue helping in the future (beyond my due date). So, I consider December 18th as a deadline for preliminary tests.
The idea of this package, at least for me, came because I was implementing a RDD using CCT bandwidth but the software I was using was not able to handle the number of observations. Said so, I'm not attached to the idea of implementing CCT. I think implementing the McCrary test works perfect for me too. Specially if that helps the package and I don't overlap with @nicolas-jimenez 's work.
Thank you for your thoughts. I need to read more about package licensing.
Best,
Pablo

[nignatiadis]

Hi Pablo,

This sounds great! Let me know if there is any guidance I can provide.

Best,
Nikos

[nicolas-jimenez]

Hi Pablo and Nikos,

Apologies for the delayed response. I am in the process of applying to PhD programs now, so I am somewhat time constrained for a couple of weeks and then plan to begin working on this.

Given this, Pablo, if you prefer to start working on CCT now, that is fine with me. I'd be happy to plug in in a few weeks once I've submitted most applications. Whichever you prefer, I'll stay tuned.

Best,
Nicolas

[pvalenzuelac]

Hi @nicolas-jimenez
I will start working on the McCrary test first. After that, we can talk for CCT.
Good luck with the applications!

[nicolas-jimenez]

Perfect, sounds great.
And thanks very much!

[nignatiadis]

Good luck with your applications @nicolas-jimenez!

[nicolas-jimenez]

Thank you!!