Discrepancies in results [message #24589] |
Sat, 04 June 2022 06:50 |
Mason92
Messages: 33 Registered: December 2021
|
Member |
|
|
Hello,
For Pakistan DHS 2012-13 and 2017-18, the proportions for emotional and physical violence reported in the country reports are different from the ones I am getting, by provinces, as well as by provincial urban/rural residency status.
Please see attached the table where, discrepancies are specified in red color, between the results I got, and the ones reported in country reports; Pakistan DHS 2012-13, table 14.6 on page 228, and for Pakistan DHS 2017-18, table 16.10 on page 319.
I did use the correct survey weights i.e. [d005 / 1000000].
Please guide.
Thanks
|
|
|
|
Re: Discrepancies in results [message #24608 is a reply to message #24603] |
Tue, 07 June 2022 12:39 |
Mason92
Messages: 33 Registered: December 2021
|
Member |
|
|
I am using Stata; for Pakistan DHS 2017-18, most (not all though) of my results match exactly with the ones provided in the country report. But for Pakistan DHS 2012-13 there are lots of discrepancies.
My code is attached.
Kindly guide.
Thanks
|
|
|
Re: Discrepancies in results [message #24643 is a reply to message #24608] |
Mon, 13 June 2022 11:08 |
Janet-DHS
Messages: 880 Registered: April 2022
|
Senior Member |
|
|
Following is a response from Senior DHS Specialist, Kerry MacQuarrie:
The issue is a very small one. In 2012-13 Pakistan DHS, the "ever experienced emotional violence by spouse" and "ever experienced physical violence by spouse" variables have 1-2 cases that are coded as 9 (declined to answer). See below:
In the DHS construction of the variable, these cases are included in the denominator, but not the numerator. In your code, these cases are excluded from both the numerator and denominator (set to missing). If you recode them as "no" (ie, include them in the denominator), the results match those in the DHS final report tables. The following code does exactly that and works for both survey years:
*emotional violence
recode d104 (9=0),g(ev)
*physical violence
gen pv=d106
replace pv=1 if d107==1
recode pv(9=0)
The discrepancy you are getting in the 2017-18 survey results is a matter of the difference in the way Stata versus the program DHS uses to run Final Report tabulations handles rounding, e.g. 21.25 will get rounded down to 21.2 instead of rounded up to 21.3.
|
|
|
|
Re: Discrepancies in results [message #24708 is a reply to message #24643] |
Sat, 25 June 2022 14:00 |
Mason92
Messages: 33 Registered: December 2021
|
Member |
|
|
Thanks again, but I notice that for 2012-13 Pakistan DHS, the urban and rural provincial indices for intimate partner violence are flipped. I am getting exactly the same proportions for emotional and physical violence by provincial urban rural residency status, but my urban proportions match with the rural proportions and vice versa, based on variable v023 coding.
Please guide.
Thanks
|
|
|
|
|