RD plot command: -rd-, -rdplot-, -cmogram-

[Alessandro Sovera]

Good evening,

I am rurnning a RD design for my Master thesis and I am trying to plot the most intuitive (and pretty) graphs of RD effect.

I have seen from past posts on this forum and googling around that there are at least three commands (ssc) that do the job: -rd-, -rdplot- and -cmogram-. The issue I am stucking against is the fact that each command has something missing: -rdplot-, which seemed to me the most up-to-date lack confidence intervals (as far as I know); -rd- gives me a strange error message and compute a bandwidth that is clearly wrong; -cmogram- does not permit an intuitive representation of the results.

I post here my code examples

*-rd-
rd educ_years T run_var, z0(0) gr bin bin(educ_years) ox k(rectangle)

*-rdplot-
rdplot educ_years run_var, c(0) binselect(espr)

*-cmogram-
cmogram educ_years run_var, cut(0) ci(95) l(0) lowess scatter

Is anybody of you experienced with this commands and willing to help me chosing the best one?

Thank you

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, Stata output, and sample data using dataex.

It looks like you're using a user written program. I'd double check that a conventional Stata procedure doesn't do the job as well. [Stata 15 does many things we used to need user written programs to execute). Sometimes very few of us have even downloaded a specific use-written program let alone used it enough to provide much help. I also believe Stata tests their routines more rigorously than most authors of user-written routines so they're less likely to have glitches.

You might want to try to tell us what you're trying to estimate and the kind of data you have. The way you've pose your question, we can't help you even if we know a good routine for your problem unless we know what rd does.

[Alessandro Sovera]

Ok, sorry I supposed the commands to be known since I found some other questions about the topic. My question was direct to someone having familiarity with this kind of analysis/commands that I suppose there are since I found some related (but not sufficient) discussion on this same forum.

I am running a program evaluation using a regression discontinuity design so as a first step I wish to plot the rd effect by plotting the years of education of individuals against the running variable that in my case is years. The idea is that I should find a discontinuity in the cutoff year, i.e. the year in which policy has been implemented.

It is almost impossible to show an example and so I just reported (as requested in the FAQ) my code in code delimiters for the three commands I found should do the job, i.e. estimating two fitted lines before and after the cutoff in order to check the existence of any discontinuity along time. I wasn't talking about glitches, it's just that this commands are not complete in themselves. In my opinion, a complete command would plot the lines together with confidence intervals and in a tidy way. Unfortunately, I am working on Stata 14 using my university licence so I have done what I could.