| Domestic violence [message #19046] |
Wed, 08 April 2020 07:43  |
Catherine K
Messages: 4
Registered: April 2020
|
Member |
|
|
Hello all,
I am conducting research on IPV against men in Kenya using DHS 2014 using approx 10 variables both demographic and on physical & sexual abuse by former/current wife or partner.
When i run statistical analysis, i get a lot of missing values. How do i focus mainly on men who participated in the DV interview (only 39%) (SMDV) and how do i track/flag who answered the related demographic questions.
In addition:
MV501
MV502
MV504
All address the former/current partner. Data did not make sense when I recoded then combined.
MD106, 107, & 130A all address physical IPV.
MD108 & MD 130B- sexual abuse. Can I combine them?
MD120- is blank, no data
Thanks.
Catherine.
|
|
|
|
| Re: Domestic violence [message #19058 is a reply to message #19046] |
Fri, 10 April 2020 10:59  |
Trevor-DHS
Messages: 803
Registered: January 2013
|
Senior Member |
|
|
The variable SMDV indicates whether a man was selected for the DV module in this survey. Select for SMDV equal to 1. The demographic variables MV501 and MV502 should be available for all cases. MV504 only has data for those reported as married or living together in MV501. MV502 is just a recoding of MV501. You say that data did not make sense when you combined them, but this does not give enough information. Please provide details including your code and output for us to be able to help.
For the MD variables you listed, again, your explanation is insufficient to be able to help. Please provide more details of the issues you are having together with your code and output.
MD120 is blank as the question was not asked in the Kenya survey.
[Updated on: Fri, 10 April 2020 11:00] Report message to a moderator |
|
|
|
|
|
|
|
|
|
| Re: Domestic violence [message #19490 is a reply to message #19464] |
Tue, 30 June 2020 17:03 |
Bridgette-DHS
Messages: 3199
Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
When you combine two variables to construct a third, it's better to describe that as a recode rather than a merge. You have not given the recode commands, so I cannot interpret the different values. The most obvious recode of two variables, each of which is coded 0 for no or 1 for yes would be into three categories: 0: no/no; 1: yes/no or no/yes; 2: yes/yes
It's up to you whether combining these two variables into one is meaningful. But if you do construct a recode like this I strongly recommend that you treat the combinations as ordered categories and not use a method such as linear regression that assumes they are equidistant.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
