Two way plot with RD design

[abhinav khemka]

Hi,

I am running an RDD regression with covariates and clustered standard errors using rdrobust package in Stata/SE 16.1. I wanted to plot the RD estimates using confidence intervals as lines. I have tried reproducing the rdplot using a twoway graph and it works well. I am trying to restrict the yscale for better visualization of the plot, but none of the usual options work. Any help will be much appreciated!

Code to reproduce the graph:

rdrobust y x, c(0) vce(cluster cluster_id) covs(yr1-yr9)  

local bw=e(h_l)
local doublebw=(e(h_l)*2)
local halfbw=(e(h_l)/2)
local cct_n = e(N_h_l)+e(N_h_r)

rdplot y x if inrange(x, -`bw',`bw'), c(0) kernel(triangular) genvars nbins(20 20) covs(yr1-yr9) vce(cluster cluster_id) p(1) hide  

tw (scatter rdplot_mean_y rdplot_mean_x ) ///    
(lfitci y x if x < 0 & inrange(x, -`bw', `bw') , ciplot(rline)) ///    
(lfitci y x if  x>= 0 & inrange(x, -`bw', `bw'), ciplot(rline)), ///  
 xline(0)

[Nick Cox]

Cross-posted on Stack Overflow at https://stackoverflow.com/questions/...esign-in-stata

Here is what I said there

That's not code to reproduce the graph without the variables concerned. I can't see what code you are trying to restrict the yscale, but it's documented that ysc() never reduces what is shown; at most it extends the axis. What you're talking about seems to mean chopping some of what Stata would show and that's not trivial without extensive coding. It's usually a bad idea statistically too. I am guessing because your post doesn't constitute a precise problem yet. Sorry if that is disappointing

Please note our longstanding request that you tell us about cross-posting. https://www.statalist.org/forums/help#crossposting

EDIT: Stack Overflow version now deleted.