Egypt 2008 DHS: IR data file versus OR data file [message #3862] |
Wed, 25 February 2015 15:53 |
catherine3h
Messages: 10 Registered: February 2015 Location: United States
|
Member |
|
|
Hello all,
I am conducting an analysis using data from the 2005 Egypt DHS and the 2008 Egypt DHS. I want to look at overall demographics and the indicators that cover HIV/AIDS and Hepatitis C during these two years. I noticed that the IR file from 2008 (N=16,527) contains the information from the women's questionnaire. It contains some health information, but no indicators related to HIV/AIDS and Hepatitis C. In contrast, the OD file from 2008 (N=12,008) contains information from men and women, and it covers demographic information as well as information on HIV/AIDS and Hep. C.
Why doesn't the 2008 IR file contain information on HIV/AIDS and Hep. C? Ideally wouldn't I need this information to compare the responses from the sample of women in 2005 to the sample of women in 2008? I could stratify the 2008 OD file by gender, but this would greatly reduce the sample size.
Any help or insight on this would be greatly appreciated.
Thank you.
Catherine
|
|
|
Re: Egypt 2008 DHS: IR data file versus OR data file [message #3951 is a reply to message #3862] |
Tue, 10 March 2015 07:41 |
Bridgette-DHS
Messages: 3203 Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS Specialist, Noureddine Abderrahim:
Hepatitis C questionnaire was administered to a subsample of the Egypt 2008 survey. All eligible men and women age 15-59 were included in this subsample. A man's survey was not conducted in Egypt 2008 only ever-married women 15-49 were interviewed. Linking the individual data will reduce the sample of person's to only these ever-married women. But, you are still able to link ever-married women to their corresponding information in the health questionnaire using V001 (Cluster number), V002 (Household number), and V003 (Woman's line number) and their corresponding variables HPSU, HNUMBER, and WLINE.
However, statistically I don't think that using either weights, the subsample weights or the ever-married sample, would be correct since we're excluding from both samples a number of respondents.
|
|
|