rdplot does not match rdrobust output

[Eamon Stocker]

Hi all,

I'm aware that there are already posts about this but none have come up with an answer. I am running rdrobust and using option 'all' to obtain robust discontinuity estimates as follows:

rdrobust recontest_2024_winners adjusted_margin_of_victory if margin_of_victory > 0 & margin_of_victory != ., c(0) p(1) all

I am attempting to replicate this visually using rdplot:

rdplot recontest_2024_winners adjusted_margin_of_victory if margin_of_victory > 0 & adjusted_margin_of_victory > -0.015 & adjusted_margin_of_victory < 0.015, c(0) p(1) h(0.015) kernel(triangular)

However, rdplot is currently only modelling the conventional estimates rather than bias-corrected, heteroskedasticity-consistent robust estimates that are outputted by rdrobust. How can I faithfully replicate the rdrobust output in rdplot?

Any help would be much appreciated

[Sebastian Calonico]

Hi Eamon,

rdplot does not compute anything related to inference. You could replicate the rd estimates by using the same bandwidth and kernel, as you are doing.

For bias-corrected estimates, first run rdrobust (setting rho=1), and rdplot with the same bandwidth, but setting p=2.

I hope this helps,
Sebastian.