callagain command is not working

[Arto Arman]

Hi Everyone

I am trying to estimate bounds for my treatment effect with Behaghel et al.'s command "callagain" (which can be found here). For it to run one needs the following:

• dependent variable y: in my case it is a standardized index variable called; laoindex
• treatment variable Z: in my case this is just a binary treatment variable; treatment
• response variable R: in my case this is a binary variable that indicate nonmissingness; nonmissing
• number of calls variable N: in my case this is how many reminders were sent out {0,1,2}; reminder

The command looks the following way:
callagain laoindex treatment nonmissing reminder, ytype("cont") graph

However, whenever I run this command, Stata tells me:
"variable Z not found"



I also tried to run it on another computer and with a Stata-provided example data set, leaving me with the same error.
Did anyone encounter a similar problem with this command? If so, were you able to fix it?

Thanks a lot for the help!
Best,
Arto

* Example generated by -dataex-. For more info, type help dataex
clear
input float(laoindex treatment nonmissing reminder)
  -.2879483 0 1 2
 -1.9779216 0 1 2
 -1.5065705 0 1 0
  1.1951087 0 1 0
 -.09251682 0 1 0
 -.11548679 1 1 1
          . 0 0 2
          . 1 0 2
   1.746948 1 1 0
 -.56386805 0 1 0
          . 0 0 2
-.012028575 0 1 1
 -1.0352192 0 1 0
          . 1 1 2
  -.2879483 0 1 0
   -.575353 1 1 0
 -1.5065705 1 1 1
          . 0 1 2
          . 1 1 2
  -.4718948 1 1 1
  2.2182992 0 1 0
          . 1 0 2
          . 0 0 2
  1.1951087 1 1 0
   1.275597 0 1 0
 -1.0352192 1 1 0
  1.5630016 0 1 0
  -1.322624 0 1 0
  -2.449273 1 1 1
 -1.0352192 0 1 1
          . 0 0 2
          . 0 0 2
  2.2182992 0 1 2
          . 0 0 2
   1.379055 0 1 0
 -.11548679 1 1 2
 -.58683795 0 1 0
  -2.449273 1 1 1
          . 0 0 2
   1.746948 0 1 0
   1.746948 0 1 0
          . 0 0 2
   .9077039 0 1 1
          . 1 1 2
  -.2994333 1 1 0
   .5398109 0 1 0
  -.2994333 0 1 2
          . 0 0 2
   1.746948 0 1 0
   .6317841 0 1 0
   1.379055 1 1 0
          . 0 0 2
          . 0 0 2
          . 0 0 2
          . 0 1 2
          . 0 0 2
 -.58683795 1 1 0
   .6432691 0 1 0
          . 0 0 2
          . 0 0 2
          . 1 0 2
          . 0 0 2
  2.2182992 0 1 1
          . 0 0 2
  1.5630016 0 1 0
 -1.9779216 0 1 0
 -1.5065705 1 1 2
          . 0 0 2
   .3558644 0 1 1
 -.09251682 0 1 2
  2.2182992 0 1 0
 -1.0352192 0 1 0
          . 0 0 2
          . 0 0 2
          . 0 0 2
          . 1 0 2
          . 0 0 2
          . 0 0 2
 -1.0352192 0 1 2
          . 1 0 2
 -1.6100286 0 1 0
  -.7707845 0 1 1
          . . 1 .
          . 1 0 2
          . 1 0 2
   1.746948 1 1 0
          . 0 0 2
          . 0 0 2
          . 0 0 2
   .7237574 1 1 0
 -1.5065705 1 1 1
  2.2182992 0 1 1
  .25240618 1 1 0
          . 0 1 2
          . 0 0 2
          . 0 0 2
 -.20746003 1 1 0
   -.575353 1 1 0
          . 0 1 2
          . 0 0 2
end

[Nick Cox]

It's a bug. The authors have a line

qui sum Z if `touse' [`weight'`exp']

which fails as you noticed. They probably intended

qui sum `Z' if `touse' [`weight'`exp']

While you're fixing that it would be better as

sum Z if `touse' [`weight'`exp'], meanonly

There may be other bugs.

[Nick Cox]

That last line in #2 should be

 
 sum `Z' if `touse' [`weight'`exp'], meanonly

[Antonio Ponce]

Hello,

I have the same problem. Arto Arman did you figure it out?

Thanks!

[Nick Cox]

Antonio Ponce Why refer to the OP when you have suggestions about fixes?

[Antonio Ponce]

Hello!

I finally figured out how to make the command work. The solution was to rename my variables with the same notations that the authors use. Therefore, what will work for you is:

rename laoindex Y
rename treatment Z
rename nonmissing R
rename reminder N

Good luck!

[Arto Arman]

Antonio Ponce Nick Cox Thank you guys so much! It's working now!

[Marouchka Viola]

Hello,
I am trying to work with this command. Nevertheless, it is not working even when I am renaming my variables. Here is my code :

rename bribe Y
rename treatment Z
rename nonmissing R
rename tentative N

callagain Y Z R N, ytype("bin")

I obtain this error : invalid syntax

Did you modify anything in the ado file ?


Thanks @Antonio Ponce @Nick Cox Arto Arman for your previous messages.