re: st: -atts- does not produce standard errors in some cases

["Ariel Linden" <ariel.linden@gmail.com>]

Hi Ryan,

Have you checked the individual strata (blocks) to see if there are at least
2 treated units? Perhaps there is only one treated unit in those strata with
missing SE? So you might get a mean but no SE. As you say, you didn't
provide sufficient info.

I'd start by checking the values in each strata manually, -tab treat block,
sum(outcome)- 

Ariel

Date: Tue, 4 Feb 2014 15:58:15 -0500
From: Ryan Turner <rjturner@cmu.edu>
Subject: st: -atts- does not produce standard errors in some cases

Dear Statalist,

I am using -pscore-, -atts-, and -attnd- to estimate a large number of
propensity score models (Becker 2002).  In some rare cases (a relatively
small subset of my total model), -atts- (ATT stratification matching)
produces an estimate for ATT, but the standard error is missing (.).
-attnd- on these same models produces valid (not missing) standard errors.

What could cause this?  The missing SEs are well distributed across my model
types, so I can't directly attribute it to my model (e.g., using an
all-dummies pscore specification in some cases).  My only thought is that
maybe I could try bootstrapping my standard errors.

I recognize I haven't given enough information to fully answer this question
(e.g., a minimum reproducible example); my goal is simply to understand what
might be happening, so any thoughts are appreciated.

Best,
Ryan

References:
Becker, S. O., & Ichino, A. (2002). Estimation of average treatment effects
based on propensity scores. The Stata Journal, 2(4), 358-377.

- --
Ryan J. Turner <rjturner@cmu.edu>
Engineering and Public Policy
Carnegie Mellon University
+1-412-304-5014 (C) | +1-484-483-3244 (GV)


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/