There are ~130 RD questions on the site (with some false positives) and it is a standard quasi-experimental inference method in economics and other fields. Does this volume merit its own tag?
$\begingroup$
$\endgroup$
6
1 Answer
$\begingroup$
$\endgroup$
6
I don't see why not. That is a perfectly valid topic. We have other, similar tags for methods used to infer causality from observational data, including:
- causality (387 threads)
- granger-causality (137 threads)
- instrumental-variables (292 threads)
- propensity-scores (201 threads)
- difference-in-difference (187 threads)
- fixed-effects-model (511 threads)
- etc.
As always, if you do create the tag, please add a good excerpt to guide usage, and preferably a good full wiki as well.
It seems we already have a discontinuity tag with 40 threads (ht, @amoeba). Of those, 22 have both [regression] and [discontinuity] and certainly more than half are really about RD. There are also 17 threads with [regression] and "discontinuity", but without the tag. I think it would be better to have an explicit [r-d] tag specifically for RD. The term / idea of discontinuity is somewhat generic. It could cover RD, but could also be used for something like changepoint or something else (e.g., Is there a way to correct for continuity for the Shapiro-Wilk test?). If we did create a new tag for RD, the question then is what to do w/ [d]. My first thought would be to go through those 40 threads and properly retag them, leaving [d] only on those threads like the one just linked that are about other senses of discontinuity and that can't be given a more specific tag. Then we could edit the wiki excerpt to provide better usage guidance.
answered Oct 19, 2017 at 15:04
gung - Reinstate MonicaModgung - Reinstate Monica
147k5 gold badges79 silver badges136 bronze badges
-
$\begingroup$ Agreed, regression-discontinuity is a relevant methodology in the Econometrics literature. $\endgroup$ Commented Oct 22, 2017 at 22:20
-
1$\begingroup$ I upvoted several days ago, but now noticed that there exists discontinuity tag with 40 Qs. One would need to figure out what to do with it before creating [regression-discontinuity]. $\endgroup$– amoebaCommented Oct 24, 2017 at 11:16
-
$\begingroup$ Re Update: I agree but I'd say that in that case it would be better to eliminate [d] altogether. I am not sure there is any useful (useful for a tag) meaning of "discontinuity" once we take [r-d] out of it. $\endgroup$– amoebaCommented Oct 24, 2017 at 13:27
-
$\begingroup$ That's a viable option, @amoeba. I'm open to that. This was just my first thought. $\endgroup$ Commented Oct 24, 2017 at 13:43
-
$\begingroup$ Personally, I prefer the [regression discontinuity] rather than [r-d] or [rdd]. I think there is still room for [discontinuity]. There seem to be some questions about detecting structural breaks and their ilk, and that seems qualitatively different that making use of one that is known to exist. $\endgroup$– dimitriyCommented Oct 25, 2017 at 1:02
-
$\begingroup$ @DimitriyV.Masterov, "r-d" & "rdd" is just my shorthand b/c I'm too lazy to type "regression-discontinuity" each time. But I mean that the new tag should be
[regression-discontinuity]. $\endgroup$ Commented Oct 25, 2017 at 1:15
[discontinuity]tag? Should we initiate another thread about that? (cc, @amoeba) $\endgroup$