| Estimation of level-weights using the Couple-Recode (CR) data from the DHS dataset [message #27057] |
Fri, 16 June 2023 18:53  |
DHS user
Messages: 111
Registered: February 2013
|
Senior Member |
|
|
I am presently writing a research paper titled' Household economic consequences of Nigerian married or in-union women of reproductive ages meeting their need for family planning with modern contraceptives (MC)' I plan on using the couples recode (CR data) since I need information on couples so that variables such as the educational level of household head, or the employment status or assets ownership status of the household head, and age of marriage of the household head, and so on, can be obtained.
In relation to this, I cannot use the household recode data because information on women with a need for family planning, and their subsequent status of either meeting this need or not with MC, is not available therein. On the other hand, although the individual women's recode data contain information on the need for family planning, I cannot obtain detailed information about the household head's characteristics in situations where the household head is male, as is mostly the case in developing countries such as Nigeria.
I plan on carrying out a multilevel analysis, with the couples nested within clusters. The latter level will have variables such as place of residence(whether rural or urban), cluster polygyny status, and so on. In relation to this, therefore, I have to calculate level weights. In the paper titled 'Multilevel Modeling Using DHS Surveys: A Framework to Approximate Level-Weights', in which you were a co-author, the first step in approximating level-weights for use in the multilevel analysis was said to be the de-normalization of final survey weights using an approximation method. In relation to this my question is what should I use as the final weight given that the CR data has two final weights; that for the individual women and then that for the individual men?
Emomine
[Updated on: Fri, 16 June 2023 18:54] Report message to a moderator |
|
|
|
| Re: Estimation of level-weights using the Couple-Recode (CR) data from the DHS dataset [message #27058 is a reply to message #27057] |
Fri, 16 June 2023 18:56  |
DHS user
Messages: 111
Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS staff member, Tom Pullum:
Our general practice is to use the men's weight (mv005) for the CR file. That's because men's nonresponse is usually more serious than women's nonresponse.
You cannot assume that the male partner is the household head. Some households will be more complicated than that. I propose the following lines in Stata. If you do this, you will not need to use the CR file at all.
Not sure this is what you want, but it's an option if you want to pair the woman with the household head. Note that it can happen that the woman is the household head. This will be the case if v003=head_hvidx. Cheers--Tom
* Construct a file that attaches the characteristics of the household head to each
* eligible woman in the household
* June 16, 2023
* specify a workspace
cd e:\DHS\DHS_data\scratch
use "C:\Users\26216\ICF\Analysis - Shared Resources\Data\DHSdata\NGPR7BFL.DTA", clear
* reduce to just household heads
keep if hv101==1
* select variables you need, not necessarily as illustrated here
keep hvidx hv0* hv1*
* rename to avoid possible confusion
rename hv* head_hv*
* specify variables for the merge
rename head_hv001 cluster
rename head_hv002 hh
sort cluster hh
save temp.dta, replace
use "C:\Users\26216\ICF\Analysis - Shared Resources\Data\DHSdata\NGIR7BFL.DTA", clear
rename v001 cluster
rename v002 hh
sort cluster hh
* merge is m:1 because there may be more than 1 eligible woman per household
merge m:1 cluster hh using temp.dta
tab _merge
* _merge=2 for households that do not include any eligible women
drop if _merege==2
drop _merge
[Updated on: Fri, 16 June 2023 18:58] by Moderator Report message to a moderator |
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
