Announcement

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

  • Propensity score matching and ATT analysis for multiple treatments

    Hello,

    I want to analyse the ATT of an intervention using PSM. My treatment has three arms: the control group, treatment 1 receiving only treatment 1, and treatment 2 receiving both treatment 1 and 2. In a first step, I simply compared receiving any treatment to being in the control group using the following commands:
    Code:
    probit treated cov1 cov2 cov3
    predict pscore_treated
    
    foreach y of varlist dep1 dep2 dep3 {
    psmatch2 treated, outcome(`y') pscore(pscore_treated) kernel common 
    eststo nn_treated_`y'
    esttab nn_pooled_dep1 nn_pooled_dep2 nn_pooled_dep3, keep(_treated)
    }
    Where treated is a binary variable indicating if any treatment was received, cov1-3 are the covariates I want to use for the matching, and dep1-3 are my different outcome variables of interest.

    Now I want to look at the effects of the two treatments separately. As far as I have understood, that requires three different comparisons: Control vs T1, Control vs T2, T1 vs T2. I already estimated the probabilities of being in the different treatments using mlogit and estimated the pscores:
    Code:
    mlogit treatments cov1 cov2 cov3
    predict pscore_control pscore_t1 pscore_t2
    However, I struggle to find how to estimate the different effects using psmatch2. Using teffects with psmatch also does not seem to be an option with multiple treatments.

    I am using Stata18.

    Thank you!

  • #2
    teffects ipwra allows for multinomial treatment

    Comment


    • #3
      Agree with George. IPWRA is a good way to go. I suppose you can do each treatment level separately using the controls to match. This likely will be less efficient than IPWRA.

      Comment


      • #4
        Thank you so much George and Jeff. Using the teffects ipwra command works perfectly to estimate ATTs for multivalued treatments!

        Comment

        Working...
        X