The DHS Program User Forum: Weighting data » Variable Verification in Zim DHS 2010-11

Home » Data » Weighting data » Variable Verification in Zim DHS 2010-11

Show: Today's Messages :: Show Polls :: Message Navigator

Variable Verification in Zim DHS 2010-11 [message #2792]

Mon, 25 August 2014 18:35

swdalb
Messages: 5
Registered: August 2014

Member

Hi All,

I think that the data in the women's questionnaire for the 436 and 437 variables (m62 and m66) variables in the Zim 2010-11 DHS dataset at the birth level have been switched. I.e. the recode manual says that 436 - did anyone check up on you before you left the facility - is supposed to be recoded as m62 and that 437 - did anyone check up on you after you left the facility - should be recoded as m66, but they are reversed in the actual data. I figured this out by checking some skip patterns in the questionnaire. For example, the qx says that "No" responses on 437 / m66 should skip down to 446. So, 442 - did the baby receive a postnatal checkup within 2 months of birth (recoded as m70) - should have all missing values for instances where m66 == 0. However, if you run tab m70 if m66 == 0,m you get a number of different responses BUT if you run tab m70 if m62 == 0,m you get all missing values, as you should get if m62 was actually variable 437, not 436. Could someone either confirm that these variables are switched for me or show me the error of my ways? :) Thanks a ton!

-Sean

Report message to a moderator

Re: Variable Verification in Zim DHS 2010-11 [message #2795 is a reply to message #2792]

Mon, 25 August 2014 20:24

Trevor-DHS
Messages: 805
Registered: January 2013

Senior Member

Hi Sean,

I checked the recoding and it is done correctly. I think your confusion stems from two reasons, though:
1) m66 is created not only from q437, but also from q438.
2) there is an additional skip in 441 that skips those saying yes in q437 to q446.
The result of this latter skip is that those saying either "yes" or "no" on q437 both ultimately skip to q446, which means that all of those saying "no" on q436 skip over q442 to q446. Thus you should find all cases code "no" on q436 (m62) are skipped (missing) on q442 (m70).

Report message to a moderator

Re: Variable Verification in Zim DHS 2010-11 [message #2797 is a reply to message #2792]

Mon, 25 August 2014 21:22

swdalb
Messages: 5
Registered: August 2014

Member

Ok thanks Trevor. Can I ask you to verify some code for me then? This skip pattern still seems bizarre to me and I'm not sure I'm generating the variables I want correctly. These are the variables I need:

* Any postpartum care (at any time)
* Postpartum care within 2 months of the delivery
* Postpartum care within 2 days of the delivery
* Postpartum care by a skilled provided (doctor, nurse, midwife)
* Any postnatal care (at any time)
* Postnatal care within 2 months of delivery
* Postnatal care within 2 days of delivery
* Postnatal care by skilled provider
* Postnatal care in a facility (anything but their home or another private home).

I've attached my code so far and changed it based on what you said below. My main difficulty is that I'm trying to give values of 0 to observations that were skipped based on the qx logic, but it doesn't look like I should, for example, give the "any PNC" variable a 0 simply because it was skipped over. Anyway, let me know what you think and thanks again - this was really helpful.

-Sean

Attachment: ppc_pnc_var_construction.do
(Size: 1.48KB, Downloaded 939 times)

Report message to a moderator

Re: Variable Verification in Zim DHS 2010-11 [message #2811 is a reply to message #2797]

Tue, 26 August 2014 13:13

swdalb
Messages: 5
Registered: August 2014

Member

Hey again Trevor,

So, I think I could do everything if I saw how w436 through w438 get translated into the recode. I really need a variable that JUST has w437 in it to do what I want properly. I should be good to go if I can reconstruct that accurately. Thanks!

-Sean

Report message to a moderator

Previous Topic:	CS plan
Next Topic:	Svysey vs iweght

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Mon Dec 30 11:01:46 Coordinated Universal Time 2024