|Weighting Data for Pooled Indonesian DHS Dataset [message #10033]
||Fri, 17 June 2016 00:07
Registered: May 2016
I have several questions regarding the weighting and survey set of Indonesia DHS dataset 2007 and 2012
So currently I want to pooled the BR dataset for both period and would like to generate pooled wealth index using PCA in order to examine the change in wealth between the years, in which I need to weight the data. I'm using STATA for the analysis, and from what I understand I must do the following before I merge the dataset
1. De-normalizing the sample weight
gen wgt_denorm=wgt*total number of female population age 15-49 in the particular period / number of female age 15-49 interviewed in the survey
My question about this part is whether it is necessary to use the total number of female population age 15-49 in the particular period or it is sufficient to do the following command?
quietly summarize v005
If it is indeed necessary to use the total number of female population age 15-49 in the particular period, will it be okay for me to estimate this number by multiplying the total country population projected for that particular period with the percentage of female age 15-49 in the household population as reported in the DHS final report on that year?
2. Correcting the PSU and strata to make it specific for the particular year e.g. by adding 10000 for PSU and strata in 2007 and 20000 for PSU and strata in 2012
Question for this part, I generate the strata manually by using the following command
egen strata=group(v024 v025), label
Is this the correct strata to use for Indonesia DHS 2007 and 2012? Or should I use v022 instead?
3. Setting the survey design (after merging the dataset) with the following
svyset [pweight=wgt], psu(v021) strata(strata)
Are these steps correct and do I need additional steps to correct the weighting?
Thank you very much beforehand