| Difference in N when HIV data merged [message #18823] |
Wed, 26 February 2020 10:09 |
behayes4
Messages: 24
Registered: February 2020
|
Member |
|
|
Thank you. I'm going through systematically to merge country HIV data with men and women datasets. The biggest issue I've caught so far is Burundi 2016-2017. There are 21,955 people in the HIV dataset. When I match and merge in men and women datasets, I am getting 15,904 who have HIV data. I cannot identify why this drop off happens.
Here is how I am recoding for the initial merge to match HIV data:
recode mv001 (else=copy) into HIVCLUST.
recode mv002 (else=copy) into HIVNUMB.
recode mv003 (else=copy) into HIVLINE.
EXECUTE.
recode v001 (else=copy) into HIVCLUST.
recode v002 (else=copy) into HIVNUMB.
recode v003 (else=copy) into HIVLINE.
EXECUTE.
I am then doing a one to many merge (Syntax is country specific but here is an example).
DATASET ACTIVATE DataSet40.
SORT CASES BY HIVCLUST HIVNUMB HIVLINE HIV03 HIV05.
DATASET ACTIVATE DataSet39.
SORT CASES BY HIVCLUST HIVNUMB HIVLINE HIV03 HIV05.
DATASET ACTIVATE DataSet40.
MATCH FILES /FILE=*
/TABLE='DataSet39'
/BY HIVCLUST HIVNUMB HIVLINE HIV03 HIV05.
EXECUTE.
I am then simply tracking the Ns in an Excel file sheet to identify where missing data appears. So far the other 4 nations I coded have about a 2-5% dropoff between HIV full dataset and then the Ns reported when the data is merged.
Anyone else have this issue?
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
