| Re: Merging and appending Kenya DHS [message #12914 is a reply to message #12907] |
Mon, 07 August 2017 14:06   |
Bridgette-DHS
Messages: 3117
Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS Stata Specialist, Tom Pullum:
I prefer "renormalize" to "denormalize"....
When using a cross-sectional analysis (e.g. for 2014) you do not need to renormalize, but you do need to choose between weights. If any variables in a regression or tabulation, etc., come from the IR file I would use v005 rather than hv005. v005 is equal to hv005, except for a slight adjustment for loss of a few women respondents. Similarly, if you are using variables from the MR file, then use mv005 rather than hv005. If you were using the couples file, that includes both v005 and mv005 on the same record, then preference is given to mv005 because male nonresponse is higher than female nonresponse.
I went ahead and did the merge for 2014 with the following Stata lines:
use e:\DHS\DHS_data\MR_files\KEMR70FL.dta, clear
keep mv001 mv002 mv003 mv005
rename mv001 hv001
rename mv002 hv002
rename mv003 hvidx
gen sex=1
save e:\DHS\DHS_data\scratch\KE_temp.dta, replace
use e:\DHS\DHS_data\IR_files\KEIR70FL.dta, clear
keep v001 v002 v003 v005
rename v001 hv001
rename v002 hv002
rename v003 hvidx
gen sex=2
append using e:\DHS\DHS_data\scratch\KE_temp.dta
sort hv001 hv002 hvidx
save e:\DHS\DHS_data\scratch\KE_temp.dta, replace
use e:\DHS\DHS_data\PR_files\KEPR70FL.dta, clear
keep hv001 hv002 hvidx hv005
sort hv001 hv002 hvidx
merge hv001 hv002 hvidx using e:\DHS\DHS_data\scratch\KE_temp.dta
tab _merge
keep if _merge==3
drop _merge
summarize *v005
pwcorr *v005
Here are the results from the last two lines:
I see that this survey only had a subsample of men. That may be an issue for the kind of analysis you want to do.
I would not say that you have a panel study; you have repeated cross-sections. If you are looking at changes from one survey to the next, you do not need to alter the weights. Because you have completely different men and women in each cross-section, and you are not (I hope!) trying to combine successive surveys in an additive way, then you do not need to renormalize. You definitely do not need to add up the weights.
If you still have doubts, please let me know (with an example) the sort of thing you plan to do with the combined file....
-
Attachment: v005.jpg
(Size: 31.66KB, Downloaded 1694 times)
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
