Get p-value for Matching within strata in psmatch2

[Rodrigo Badilla]

Hi all,

I wonder if exist some way to get p-value in PSM within strata in psmatch2.

I am able to calculate p-value using a naive PSM:

sysuse nlsw88, clear
psmatch2 married hours collgrad industry tenure , out(wage)

local pval: di %6.3f  = 2*ttail(`e(df_r)', abs(`r(att)'/`r(seatt)'))
di "`pval'"
0.974

But the mehod for a within strata in psmatch2.is:

g att = .
g seatt = .
egen g = group(race)

levels g, local(gr)

qui foreach j of local gr {

        psmatch2 married hours collgrad industry tenure if g==`j', out(wage)
        replace att   = r(att)   if g== `j'
        replace seatt = r(seatt) if g== `j'

}
sum att

Some tips??
Thanks in advance.
Rodrigo

[Rodrigo Badilla]

I found an alternative methods to get p-value:

First I compare 2 p-value methods using naive PSM:

sysuse nlsw88, clear

g att = .
g pvalue = .

psmatch2 married hours collgrad industry tenure , out(wage)


----------------------------------------------------------------------------------------
        Variable     Sample |    Treated     Controls   Difference         S.E.   T-stat
----------------------------+-----------------------------------------------------------
            wage  Unmatched | 7.62403932   8.16453408  -.540494767   .256021246    -2.11
                        ATT | 7.62403932   7.61148867   .012550649   .390070357     0.03
----------------------------+-----------------------------------------------------------
Note: S.E. does not take into account that the propensity score is estimated.


replace att= r(att)

// compare both methods
local pval: di %9.7f  = 2*ttail(`e(df_r)', abs(`r(att)'/`r(seatt)'))
di "`pval'"
0.9743351

replace pvalue = 2 * normal(-abs(r(att) / r(seatt)))
tabstat att pvalue, stat(mean)


   Stats |       att    pvalue
---------+--------------------
    Mean |  .0125506  .9743322
------------------------------

For PSM within strata in psmatch2:

sysuse nlsw88, clear

g att = .
g seatt = .
g pvalue = .

egen g = group(race)
levels g, local(gr)

qui foreach j of local gr {

        psmatch2 married hours collgrad industry tenure if g==`j', out(wage)
        
        replace att   = r(att)   if g== `j'
        replace seatt = r(seatt) if g== `j'
        
        * Calcular el p-value
        replace pvalue = 2 * normal(-abs(r(att) / r(seatt))) if g== `j'
}

tabstat att seatt pvalue, by(g) stat(mean)


       g |       att     seatt    pvalue
---------+------------------------------
       1 | -.2518406  .5209157  .6287709
       2 |  .3188446  .5585632  .5681149
       3 | -2.657943  2.689397  .3230036
---------+------------------------------
   Total | -.1315597  .5557906  .6094867
----------------------------------------

p-value Total: .6094867 its ok??