Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • RDD design with distances

    Hello to all,

    I am trying to apply an RRD design to a set of spatial data from rasters. The approach has been widely used in the development economics literature and I know that it is a good methodological approach. Nonetheless, the cutoff point that I had is not a "regular" one.

    As you can see in the adjunct image, the raster of distances -my treatment or running variable- has two boundaries (there are other boundaries inside but they aren't relevant), one to the left and one to the right. Both are of my interest and what I did was to calculate the distance of all pixels to the red border, the ones inside the border have negative distances values to set the cutoff at x = 0.

    What it is really worrying me is that the cutoff variable (x) is not continuous at the cutoff, at least, that is my interpretation of the density of the variable which jumps at the discontinuity. Using the rd command in both parametric and non-parametric approaches results in rare outcomes with empty Wald estimators.

    Any ideas on what it is wrong?

    Thanks in advance,

    Iván Higuera-Mendieta
    Attached Files
Working...
X