| Burundi 2010 V781 (Ever been tested for HIV) [message #8416] |
Mon, 26 October 2015 11:43  |
ODzI6g2q
Messages: 1
Registered: June 2015
|
Member |
|
|
Hi - Do you have any additional details regarding the high % of missing values to V781 for female respondents to the 2010 survey in Burundi? I have compared response by gender to this item for Burundi , Ethiopia, Kenya, Rwanda, Tanzania, and Uganda. The relative % missing for females from Burundi is striking given 9.5% with a "no" response and 47.3% with a missing item response.
Thank you!
|
|
|
|
|
|
| Re: Burundi 2010 V781 (Ever been tested for HIV) [message #8576 is a reply to message #8569] |
Tue, 17 November 2015 18:59  |
Trevor-DHS
Messages: 803
Registered: January 2013
|
Senior Member |
|
|
This is a mistake in the application that constructs the recode file. Most of the cases coded as 9 should be code 0.
The logic used in CSPro was as follows:
{ ever been tested for AIDS }
if Q916 = 1 | Q922 = 1 | Q926 = 1 then
V781 = 1
elseif Q916 = missing | Q922 = notappl | Q926 = missing then
V781 = missing
else
V781 = 0
endif;
[missing is code 9, and notappl is the equivalent of . in Stata]
in the above code, the test for Q922 should have tested for missing, not notappl.
Fortunately the information from Q922 is also stored elsewhere in the recode file in V840A, and you can correct the case as follows (in Stata):
replace v781 = 0 if v781 == 9 & v840a == . & v840!= 9
replace v781 = 9 if v781 == 0 & v840a == 9
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
