| Colombia 2010 weights [message #3014] |
Fri, 03 October 2014 15:40  |
jcuartas2
Messages: 3
Registered: October 2014
Location: Colombia
|
Member |
|
|
Hi,
I am working with the 2000 and 2010 data set for Colombia in Stata and I have a question regarding the sample weights. I am interested in knowing the total number of children who have birth registration in Colombia, so I'm using a simple tab with the weight just as described in DHS page.
With 2000 data I have no problem: the results show me there are around 4 million children under 4 years in Colombia. However, with 2010 data the results show 1.5 million children, which is not true according to Colombian census.
I realize that 2010 data for children is merged with household data, but 2000 is not. However, when I merge children and household for 2000 data set just as described in DHS page, it start to happened the same: the results show that there are like 1.5 million children in Colombia. It's like if merging children and household data sets damages the wages!
Anyone have an idea of why this is happening?
Thank you so much!
Jorge.
|
|
|
|
| Re: Colombia 2010 weights [message #3016 is a reply to message #3014] |
Fri, 03 October 2014 17:36  |
Reduced-For(u)m
Messages: 292
Registered: March 2013
|
Senior Member |
|
|
Which weights are you using on the merged data? That might be the problem. Also - do you merge household into child, or child into household? What is the unit of observation on the merged data? That might be the problem too.
|
|
|
|
| Re: Colombia 2010 weights [message #3017 is a reply to message #3016] |
Fri, 03 October 2014 17:44 |
jcuartas2
Messages: 3
Registered: October 2014
Location: Colombia
|
Member |
|
|
Thanks you so much for your kindness, today I could talk with the people in charge of the survey, and they solve my question.
Thanks!
Jorge.
|
|
|
|
| Re: Colombia 2010 weights [message #3052 is a reply to message #3017] |
Tue, 07 October 2014 14:28 |
Bridgette-DHS
Messages: 3215
Registered: February 2013
|
Senior Member |
|
|
We understand that someone already talked with you and answered your questions. However, for further clarification, I am posting the following, from DHS Senior Specialist, Guillermo Rojas:
Let me clarify a few points related to birth registration in Colombia. In 2000 as well as 2010 the birth registration was asked for children under 5 of eligible women. In DHS 2000 those data are available for living children, in variables S450A and S450B. In DHS 2010, the information is available for all children born in the last 5 years, in variables S470 and S471. This means that foster children, as well as children of mothers who have died, or who were not successfully interviewed were left out.
Regarding weights, DHS does not use expansion factors to restitute the survey data to the country's population. We use weights to compensate for subsampling or oversampling in the sampling domains of the survey. That means that our surveys produce indicators within certain confidence intervals, for the domains defined by the sample.
Nevertheless, the number of cases obtained by weighting DHS data does not expand to the country's population. Data users can therefore use the statistics produced by DHS data, and apply it to the corresponding population of the census data in the country, to obtain the desired number of cases. The variables related to birth registrations for Colombia are available in the kids data file (the unit of analysis), and therefore it should be weighted by variable V005/1000000. It is not necessary to merge the data to the household data files.
|
|
|
|
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
