Propensity score matching (PSM)

[lourdes gastearena]

Hello,
I am doing PSM in order to test the results and compare with the ones I obtained from regression analysis. I have calculated propensity scores and different methods of matching. Now, how can I assess if both models (regression or propensity score) report the same results or are different? I have problems interpreting the results.
If anyone could answer my question, I would appreciate.

Lourdes.

[FernandoRios]

Hi Lourdes
I don't think there is a straight forward answer for such comparison, because the PSM and OLS are not nested methodologies.
That being said, perhaps one alternative is to use bootstrap to test if the treatments obtained using two methods are the same.
For example:

capture program drop  comp_methods
program comp_methods, eclass
   teffects psmatch (bweight) (mbsmoke mmarried c.mage##c.mage fbaby medu)
   matrix b=e(b)
** Method2
   teffects ipwra (bweight prenatal1 mmarried mage fbaby) (mbsmoke mmarried c.mage##c.mage fbaby medu, probit)
   matrix aux=e(b)
   matrix b=b, aux[1,1]
** Method3
   reg bweight mbsmoke prenatal1 mmarried   fbaby  c.mage##c.mage   medu
   matrix aux=e(b)
   matrix b=b, aux[1,1]
   matrix colname b=m1 m2 m3
     matrix coleq b=""

   ereturn post b
  end
  
bootstrap:  comp_methods

Bootstrap replications (50)
----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5 
..................................................    50

Bootstrap results                               Number of obs     =      4,642
                                                Replications      =         50

------------------------------------------------------------------------------
             |   Observed   Bootstrap                         Normal-based
             |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
          m1 |  -210.9683   28.94586    -7.29   0.000    -267.7012   -154.2355
          m2 |  -229.9671   23.76683    -9.68   0.000    -276.5492   -183.3849
          m3 |  -222.8868   17.96025   -12.41   0.000    -258.0882   -187.6853
------------------------------------------------------------------------------
** This can be used to test equality of the predicted effects:
test  m1==m2
test  m1==m3
test  m2==m3

HTH

[lourdes gastearena]

Thank you very much!!!
It has been helpful,
Lourdes.