| Re: Peru 2004 and 2007 [message #8803 is a reply to message #5468] |
Sun, 20 December 2015 20:14   |
gyamada
Messages: 7
Registered: November 2014
Location: Baltimore
|
Member |
|
|
Hi,
I have a couple of questions about Peru 2004-2008 data (PEIR51FL) in which survey data of multiple years are in one dataset. I am analyzing mothers' BMI calculated from their weight and height data. They corrected anthropometric information in 2005 and 2007-8. My questions are:
1. Can we analyze the data separately for 2007 and 2008, or do we need to analyze in a combined manner? While the final report is available for the period of 2007-8, the method section of the report implies that the country implemented the continuous survey method in order to have estimates every year at certain levels.
2. How can we handle the sampling weights? When I calculated weighted number of observations (see the attached), the sum of weighted numbers of observations per year were different from the sum of observations for each year (although the sum of weighted number of observations across the years 2004-8 was similar to the total number of observations). In other usual surveys, the sampling weight is set so that the sum of sampling weights is similar to the number of unweighted observations.
a. More specifically, when we decide to analyze data of 2007 and 2008 combined, are we supposed to use the sampling weights as they are, or should we re-calibrate/de-normalize sampling weights for each year and combine the data of the two year?
b. Related with the above, when we want to conduct analysis on the entire dataset (e.g. difference of difference analysis), are we supposed to use the sampling weight v005 as given (of course divided by 1,000,000), or are we supposed to re-calibrate the sampling weights for each year?
Thank you,
Goro
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
