Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: pstest & psmatch2

From   Leela Hebbar <>
Subject   st: pstest & psmatch2
Date   Thu, 10 Nov 2005 12:39:26 -0800 (PST)

Can the "pstest" command be used following a
"psmatch2" command that includes the "ate" option?
Specifically, I am using the following:

/**Begin Code**/
psmatch2 itg, out(empl2_4 empl2_8 empl2_12)
noreplacement common ate;

pstest indy_g1 indy_g2 indy_g3 indy_g4 indy_g5 indy_g6
indy_g7 indy_g8 
indy_g9 educ_g1 educ_g2 educ_g3 educ_g4
pre_g1 pre_g2 pre_g3 pre_g4 pre_g5 ten_g1 ten_g2
ten_g3 ten_g4 
atl_reg so_reg no_reg qtr1 qtr2 qtr3 qtr4, sum t(itg)

/**END code**/

I gather the answer is likely no since the
documentation notes "pstest only considered balancing
for the treated" (att). However,if "pstest" can be
following a "psmatch2" with the option "ate", please
read on.  

When I run the above code with the ATE I get very odd
results, and the results  are much more reasonable
when I run the same code without the ATE.  When
invoking the ATE option, I find Pstest yields  control
that are very different than the treatment
distribution. Moreover only 3 of 219 control
individuals  (i.e. _treatment=0 & support=1) have
weight=1 and the remaining 216  folks have no value
for the weight variable. All of the  219 treatment
individuals have a weight value=1.   

In contrast, when I re-run the psmatch & pstest
command without  the ATE option, the pstest results
make much more sense and all  of the 219 control
have a weight value=1.  Also all the 219 treatment
individuals have weight=1. Is there some option I am
missing when running the routine  with ATE?  


-Leela Hebbar  

Rutgers University 
Bloustein School of Planning & Public Policy 

Yahoo! FareChase: Search multiple travel sites in one click.
*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index