Join forces?

[azev77]

@nignatiadis Thanks for this package, it looks very promising!

You're prob aware that @mdcattaneo & co already have very thorough RD code in R & STATA. It could be great to see if there are any opportunities to join forces & possibly implement a Julia version.

@mdcattaneo Julia could be an awesome language for econometrics b/c

1. Julia is almost as fast as C (way faster than R/STATA)
2. Julia has much more convenient syntax so it's easier to read/write/debug take a look @ my comparison of Julia/Matlab/R/STATA

[nignatiadis]

Hi @azev77, thank you for the suggestion and always trying to improve the Julia stats/ml/econometrics ecosystem :).

Note that this package is currently very rudimentary: the method implemented is not inferentially valid (i.e. bias is not accounted for). I was merely seeking to reproduce an empirical analysis from another paper. Unfortunately a lot of empirical work in Regression discontinuity still uses the Imbens-Kalyanaraman bandwidth selector followed by "naive" local linear regression.

I do agree that Julia provides a great language to implement state of the art approaches to estimation and inference in RD designs (such as rdrobust). The types and plot recipes in this package could be a starting point for such an effort.

[mdcattaneo]

Hi Both:

Thanks for your messages. I am not very used to communicating this way, as I do not do much social/online communication myself.

The idea of porting RDROBUST and the other packages to Julia sounds great! FYI, we are in the process of writing a grant to port our main packages to Python, and we could also add Julia there. Happy to chat more if there is interest.

Best wishes,

matias

On Thu, Oct 1, 2020 at 9:33 PM Nikos Ignatiadis notifications@github.com wrote:

Hi @azev77 https://github.com/azev77, thank you for the suggestion and always trying to improve the Julia stats/ml/econometrics ecosystem :).

Note that this package is currently very rudimentary: the method implemented is not inferentially valid (i.e. bias is not accounted for). I was merely seeking to reproduce an empirical analysis from another paper. Unfortunately a lot of empirical work in Regression discontinuity still uses the Imbens-Kalyanaraman bandwidth selector followed by "naive" local linear regression.

I do agree that Julia provides a great language to implement state of the art approaches to estimation and inference in RD designs (such as rdrobust). The types and plot recipes in this package could be a starting point for such an effort.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/nignatiadis/RegressionDiscontinuity.jl/issues/9#issuecomment-702480448, or unsubscribe https://github.com/notifications/unsubscribe-auth/AF5UBQL5HTLOD43QX3AJ5ZTSIUUVLANCNFSM4SA3VO5A .

[nicolas-jimenez]

Hi @mdcattaneo and @nignatiadis ,

I am currently a Research Fellow at MIT with Joshua Angrist and Parag Pathak. I work on projects where I have been implementing the IK and RDRobust bandwidths from Stata and R for some of our education economics projects.

I am interested in helping port the RDRobust package into Julia. @mdcattaneo - I would be happy to discuss any developments on this front. I can also follow up in an email about this.

Best, Nicolas Jimenez

[mdcattaneo]

Hi There!

Thanks for reaching out. Yes, we spoke and agreed to work on this porting next year, after we have finished some pending work on our RDROBUST package.

Please reach out back to us during the spring.

Stay safe,

matias

On Thu, Nov 19, 2020 at 4:58 PM Nicolas Jimenez notifications@github.com wrote:

Hi @mdcattaneo https://github.com/mdcattaneo and @nignatiadis https://github.com/nignatiadis ,

I am currently a Research Fellow at MIT with Joshua Angrist and Parag Pathak. I work on projects where I have been implementing the IK and RDRobust bandwidths from Stata and R for some of our education economics projects.

I am interested in helping port the RDRobust package into Julia. @mdcattaneo https://github.com/mdcattaneo - I would be happy to discuss any developments on this front. I can also follow up in an email about this.

Best, Nicolas Jimenez

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/nignatiadis/RegressionDiscontinuity.jl/issues/9#issuecomment-730662311, or unsubscribe https://github.com/notifications/unsubscribe-auth/AF5UBQP5E4FTCH55OSB2N5TSQWIGNANCNFSM4SA3VO5A .

[nicolas-jimenez]

Hi Matias,

Great, thank you for the update! I will reach back out in the Spring about this.

All the best, Nicolas