psmatch2 new variables after running the command

[Tracy Lam]

Hi,

I'm using the user-written psmatch2 command and when I run the code to implement the match a new set of variables are created, which includes _support and _weight. I have some questions about these new variables.
1. What is Stata doing when it identifies observations that are on support but their weights are missing?
2. How is _weight being calculated?
3. What does it mean to be on support?

[Melissa Garrido]

Hi Tracy,
Common support refers to the treated and comparison observations who have overlapping propensity scores and thus can be matched to each other. For instance, if you have a comparison observation with a propensity score of .02, but the range of propensity scores seen in both the treated and comparison groups is .10-.90, your comparison individual is outside the range of common support. This indicates that it has such a low likelihood of receiving treatment that it would not be a good match for any of the treated observations in your sample.
If you are doing 1:1 matching and have many more comparison individuals than treated individuals, you can have many potential matches (who are within the range of common support), but not all of them are needed. The unused potential matches are the observations on support with missing weights.
The way _weight is calculated depends on the type of matching you are doing - please post your code and output and we'll be able to help you further.
Hope this helps,
Melissa

[Tracy Lam]

Hi Melissa, this is very helpful, thanks. Here is the code and output:

. psmatch2 tx, pscore(satm2) outcome(satm3) n(2) caliper(0.05)
There are observations with identical propensity score values.
The sort order of the data could affect your results.
Make sure that the sort order is random before calling psmatch2.
----------------------------------------------------------------------------------------
        Variable     Sample |    Treated     Controls   Difference         S.E.   T-stat
----------------------------+-----------------------------------------------------------
          satm3   Unmatched | .081471322  -.094335209    .17580653   .071619083     2.45
                        ATT | .021578472   .333707657  -.312129185   .088853843    -3.51
----------------------------+-----------------------------------------------------------
Note: S.E. does not take into account that the propensity score is estimated.
 
 psmatch2: |   psmatch2: Common
 Treatment |        support
assignment | Off suppo  On suppor |     Total
-----------+----------------------+----------
 Untreated |         0        361 |       361
   Treated |        12        406 |       418
-----------+----------------------+----------
     Total |        12        767 |       779

[Melissa Garrido]

Hi Tracy,
The _weight variable for nearest neighbor matching reflects the number of times a comparison individual was used as a match.
Hope this helps,
Melissa

[Tracy Lam]

Hi Melissa,
Thanks! I do have one concern though -- when I tabulate _weight, I get some that are 0.5. Is psmatch2 doing something different here?

. tab _weight

  psmatch2: |
  weight of |
    matched |
   controls |      Freq.     Percent        Cum.
------------+-----------------------------------
         .5 |         59        8.98        8.98
          1 |        465       70.78       79.76
        1.5 |         47        7.15       86.91
          2 |         32        4.87       91.78
        2.5 |         13        1.98       93.76
          3 |         18        2.74       96.50
        3.5 |          6        0.91       97.41
          4 |          8        1.22       98.63
        4.5 |          6        0.91       99.54
          5 |          1        0.15       99.70
        5.5 |          1        0.15       99.85
          6 |          1        0.15      100.00
------------+-----------------------------------
      Total |        657      100.00