Announcement

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

  • 'rbound's after 'teffects psmatch'

    Hello everyone. Was just wondering if anyone knew if you are able to implement rosenbaum bounds sensitivity analysis using 'rbounds' after propensity score matching using 'teffects psmatch'?

    My treatment is binary my outcome variable is continuous.

    Best wishes,

    Tom

  • #2
    Unlike the psmatch2 module, teffects psmatch does not generate variables such as _support, _matched, _<outcome var>, etc. It is the latter that is necessary to run rbounds. What you could do as a workaround is the following

    teffects (outvar) (treatvar tvarlist)
    predict score, ps // generate propensity score
    psmatch2 outvar , pscore(score)
    rbounds outvar _<outvar>

    Barth

    Comment


    • #3
      A couple of corrections to the code in the previous post:

      I meant to include "psmatch" after teffects, i.e.,

      teffects psmatch (outvar) (treatvar) tvarlist).

      Second, the psmatch2 code should be:

      psmatch2 treatvar , pscore(score) outcome(outvar)

      Barth

      Comment


      • #4
        Dear Barth.

        I realised now I completely forgot to thank you for he advice.

        Thank you!

        Best wishes,

        Tom

        Comment


        • #5
          Hi Barth.

          I've been having some more trouble with this.

          In the new 'Econometric Evaluation of Socio-Economic Programs' book by Cerulli (2015), it suggests implementing rbounds by:

          gen delta = Variable - _Variable if _treat==1 & support==1

          rbounds delta, gamma (1 (1) 5)

          I was wondering how I could adapt this code to use after teffects psmatch rather than psmatch2?

          Best wishes,

          Tom

          Comment


          • #6
            Hi Tom, not sure if you figured this out yet, but here is what you could do:

            teffects psmatch ...
            predict pscore, ps
            psmatch2 treatvar, outcome(outcomevar) pscore(pscore)
            gen delta = outcomevar - _outcomevar if _treat==1 & support==1
            rbounds delta, gamma (1 (1) 5)

            Best,
            Panos

            Comment

            Working...
            X