Implementing Lee Bounds in STATA

[danishussalam]

Hi everyone,

I am using data from a randomized controlled trial to estimate lee bounds since I have differential attrition by treatment status. I am using the command leebounds which one can install via (ssc install leebounds). But I get an error saying that the selection indicator isn't specified correctly. Can someone please tell me what the issue is here?

I use the following code below. The variable "att" is the attrition variable which takes the value of 1 if the person was attrited from baseline to midline. 0 Otherwise.

. leebounds tot_pigincome_usd treat, select(att)
selection indicator att incorrectly specified
r(109);

* Example generated by -dataex-. For more info, type help dataex
clear
input double treat float(tot_pigincome_usd att)
0  57.14286 0
1       100 0
0        50 0
1 142.85715 0
0 128.57143 0
0         0 0
0         0 0
0         0 0
1  85.71429 0
1         . 1
1         0 0
1         0 0
1         0 0
0         . 1
1         . 1
1  57.14286 0
0  185.7143 0
1         0 0
1         0 0
1  71.42857 0
1        50 0
1  57.14286 0
1  64.28571 0
1         0 0
1         0 0
1         0 0
1  57.14286 0
1  84.28571 0
1  71.42857 0
1         0 0
1         0 0
1 114.28571 0
0         0 0
0  78.57143 0
1        50 0
1         0 0
1         . 1
1         0 0
1  85.71429 0
1  85.71429 0
1  28.57143 0
1 228.57143 0
0       100 0
1         0 0
0        50 0
0       100 0
1       100 0
1         0 0
0  71.42857 0
0         . 1
1  78.57143 0
1         0 0
1         0 0
1         0 0
1         0 0
1  92.85714 0
0  57.14286 0
1  78.57143 0
1         0 0
1         0 0
1  85.71429 0
0  71.42857 0
0         0 0
0  71.42857 0
1  71.42857 0
1 128.57143 0
1         0 0
1 128.57143 0
1  78.57143 0
0  71.42857 0
0  57.14286 0
0         0 0
0         0 0
0  57.14286 0
0  85.71429 0
0  57.14286 0
0  42.85714 0
0  57.14286 0
0  71.42857 0
0         0 0
0         . 1
0         0 0
1       100 0
0  42.85714 0
1  57.14286 0
0  71.42857 0
0         . 1
1  92.85714 0
1         . 1
1 142.85715 0
0         0 0
0  28.57143 0
1  64.28571 0
0  42.85714 0
0  57.14286 0
0         0 0
0  71.42857 0
0 114.28571 0
0  78.57143 0
0         0 0
end
label values treat treat
label def treat 0 "Control", modify
label def treat 1 "Treatment", modify
label values att att
label def att 0 "Retained", modify
label def att 1 "Attrited", modify

Danish

[Fei Wang]

-leebounds- assumess "att = 0" for attritted units. Switching 0s and 1s of "att" will do.

[danishussalam]

Fei Wang Thanks for the response. Just to make sure I got you correctly. You are saying I run the same command after recoding att i.e recode att (1=0) (0=1)?

It does work but I'm just trying to understand the rationale behind it.

[Fei Wang]

danishussalam , yes. You may refer to the help file for the definition of the variable in -select()-. Actually, when attrited units have missing values in the dependent variable, as in your data, you may simply drop the -select()- option.

[danishussalam]

I'll have a look. Thanks a bunch.