Home » Topics » Domestic Violence » Domestic violence var= d105a
Domestic violence var= d105a [message #23433] |
Tue, 14 September 2021 18:52 |
Jenna
Messages: 6 Registered: September 2021
|
Member |
|
|
I am using the Ghana standard DHS 2008 IR (women's) file. Looking at the variable d105a (husband ever pushed, shook, or threw something), the answers are "no", "not at all", "often", "sometimes", "9", and "yes, widow/frequency missing". What does the last answer "yes, widow/frequency missing" mean. It is a bit confusing and it is not in the DHS recode or Map files. I am not sure whether I should include it in the analysis. Although it says "frequency missing", when I tabulate the var, that specific question shows a number under frequency.
|
|
|
Re: Domestic violence var= d105a [message #23442 is a reply to message #23433] |
Thu, 16 September 2021 11:51 |
Bridgette-DHS
Messages: 3195 Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
Code 9 is not included in the value labels for the d105 variables, but in many surveys we see a sprinkling of codes such as 8 or 9 for almost any variable. These codes can be interpreted as "not stated" or something similar. I see that code 4 with label "yes, widow/frequency missing" is explicitly included in the value labels. It's hard to know what it means, but I believe "widow" would mean that the woman was not really eligible for the module and should not be included at all. "Frequency missing" probably just means that the frequency of the bad behavior was not specified.
Note that the label includes both "no" (code 0) and "not at all" (code 3). I don't have time to look at the questionnaire, but I suggest that you look and see whether these labels are correct. If they are interpreted verbatim, there is no difference between them and they should be combined, but probably there IS a difference that is not captured with the label.
How you interpret codes 4 and 9 depends on what you want to do. This is a set of 9 variables (d105a,b,c,d,e,f,h,i,j) and you may want to use them to construct a scale. If so, I suggest that you exclude, for ALL 9 variables, cases that have codes 4 or 9 on ANY of the 9 variables. In effect, recode them to ".". The impact on your results would be negligible and this approach would avoid potential inconsistencies during recoding and scale construction.
|
|
|
|
Goto Forum:
Current Time: Wed Nov 13 10:19:53 Coordinated Universal Time 2024
|