Announcement

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

  • RD analysis with covariate coefficients

    Hi Statalist

    I am performing an RD analysis investigating the effects of reduced social benefits on educational attainment among children. There seem to be some selection issues and I want to include covariates. I have used the Stata package - rdrobust - in Stata 15.1, which allows me to control for various background characteristics, but the programme does not produce covariate coefficients. I have tried - ereturn list - but rdrobust does not seem to store this information. Does anyone know how to extract the covariate coefficients - either using this package or another viable option?

    My data is confidential, but here is a piece of code using rdrobust.

    Code:
    rdrobust outcome running, p(1) kernel(triangular) h(20) covs(cov1 cov2 cov3)
    And a fictive data example
    Code:
    input float outcome running cov1 cov2 cov3)
    2 -16 1 5404 0
    3 0 3 5404 5
    5 2 3 5404 4
    1 2 1 5436 0
    8 -3 2 5404 0
    9  4 3 5438 3
    2 5 3 5436 1
    2 -1 1 5436 0
    3 -4 3 5436 3
    4 8 3 5404 4
    5 0 3 5436 3
    1 5 3 5436 3
    8 0 3 5404 5
    11 2 3 5436 3
    2  4 3 5438 3
    4 -6 3 5404 5
    9   2 3 5438 3
    1 0 3 5404 4
    1 -1 3 5404 5
    2  -6 3 5151 2
    1   2 3 5289 4
    0 -11 3 5289 2
    Anyone with a good solution?

    Thank you so much!

  • #2
    Assuming the bandwidth is optimal, what additional information would you expect from a parameter estimate for the covariates?

    Comment


    • #3
      Bandwidth is fixed to 20 in the above example. The coefficients should imply their individual effect on the effect estimate. If, for instance, gender seems to have a large influence, then this clearly indicates that my two 'experiment groups' are not balanced in terms of gender and that gender is important to the outcome.

      Comment


      • #4
        Hi Line, I am having the exact same issue, have you already managed to solve it? Thanks in advance.

        Comment

        Working...
        X