The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » Domestic Violence » Difficulty reproducing DV prevalence results
Difficulty reproducing DV prevalence results [message #589] Tue, 02 July 2013 13:37 Go to next message
mark.janko is currently offline  mark.janko
Messages: 1
Registered: July 2013
Location: Chapel Hill, NC
Member
I am trying to replicate Table 16.1 from the 2010-11 Zimbabwe DHS final report (page 254), and am having difficulty doing so. I can replicate the "often physical violence in the last 12 months," as well as the "any physical violence in the last 12 months," but I cannot replicate "Ever" or "Sometimes in the last 12 months." I am wondering if someone can help explain why this is, and provide any suggestions as to how to fix the problem.

I am using Stata 11.2, and have defined the weighting scheme as follows:

gen wgt=d005/1000000
svyset v021 [pweight=wgt], strata(v022) vce(linearized) singleunit(missing)

keep if v044==1

I have defined the physical violence variables as follows:

**Ever:**
gen epv=cond(1,d106==1|d107==1|d115y==0,0)

**Often (last 12 months):**
gen opv=cond(1, d105a==1|d105b==1|d105c==1|d105d==1|d105e==1|d105f==1|d105j= =1|d117a==1,0)

**Sometimes (last 12 months):**
gen spv=cond(1, d105a==2|d105b==2|d105c==2|d105d==2|d105e==2|d105f==2|d105j= =2|d117a==2,0)

**Any (last 12 months):**
gen adv=cond(1,odv==1|sdv==1,0)

My crosstab code, using the wealth index as an example, is as follows:

**Ever:**
tab v190 epv [iweight=wgt], row

**Often (last 12 months):**
tab v190 opv [iweight=wgt], row

**Sometimes (last 12 months):**
tab v190 spv [iweight=wgt], row

**Any (last 12 months):**
tab v190 apv [iweight=wgt], row

As I said, my results for often and any match up well to the DHS results, but ever and sometimes do not. I have attached my results, as well as those of DHS for comparison.

On a side note, the denominator in the table is reported as 6542, but I noticed that there are 2 observations coded as "9" in the d106 and d107 variables (physical violence by husband/partner), and 13 observations coded as "9" in d117 "physical violence by non-husband/partner." Would this not mean that the actual denominator should be 6529? I have run my calculations both with and without these observations, and this does not explain the discrepancy.

Many thanks for any help that can be provided!






Mark Janko,
PhD Student, Department of Geography
Predoctoral Trainee, Carolina Population Center
Spatial Health Research Group
University of North Carolina-Chapel Hill
Re: Difficulty reproducing DV prevalence results [message #739 is a reply to message #589] Thu, 05 September 2013 19:07 Go to previous message
Liz-DHS
Messages: 1516
Registered: February 2013
Senior Member
Dear User,
Hope this will help. Using CSPro here is some code and comments for the table you are trying to run. I am using code from our standard tables, so this may not match exactly but will give you an idea of how the application works. It may assist you in replicating the results. The variables used in our tables are working variables, but you can see where they actually come from.

Row Variables (separated by + signs)
V013W refers to V013 Age in 5-year groups
v130w refers to V130 Religion
v131w refers to V131 Ethnicity
v102w refers to V102 Type of place of residence
v101w refers to V101 Region
v501w2 refers to V501 Current marital status
v218w refers to V218 Number of living children
v741w3 refers to V741 Type of earnings from respondent's work
v106w refers to V106 Highest educational level
v190w refers to V190 Wealth index
tot1549 refers to a Total Variable for age 15-49

Column Variables (come after the row variables and separated from row variables by a blank)
coldv01a refers to a Column DV01 Variable and probably refers to the first two values in your table Ever, Oftencoldv01b refers to a Colulmn DV01 Variable and probably refers to the second set of values Sometimes, Any
numwom refers to the column with the number of women

If there was a * separator, that would indicate a layer

PROC GLOBAL
{+---------------------------------------------------------- ----------------------------------+}
{+ +}
{+ +}
{+ +}
{+ CHAPTER DV. Domestic Violence +}
{+ +}
{+ Table DVWT1: WORKING TABLE Background characteristics of respondents for the DV module +}
{+ Table DV.1 : Experience of physical violence +}

set explicit;

numeric i, imax, itot, j, jmax, jtot, imar, imar1, iepv, jepv, jany, jmax1, jmax2, iage, jphs;
numeric rweight;
numeric PVever, SVever, cnt;
numeric anyPV, anySV, AnyEV, freqPV, freqSV, freqEV, decision, wbeat, diffage, diffed;
numeric chusb, fhusb;
numeric d105aw, d105bw, d105jw, d105cw, d105dw, d105ew, d105fw, d105hw, d105iw, d105kw, d103aw, d103bw, d103cw;

crosstab float(0) txxx unweight runday+runmonth+runyear
exclude(specval, rowzero, colzero, totals, percents)

title( "Tables for Domestic Violence, Country 2011" );

crosstab float(1) dvwt1 v013w+v130w+v131w+v501w2+v741w3+v102w+v101w+v106w+v190w+tot1 549 imarr*coldvwt1
exclude(rowzero,colzero,percents,totals,specval)

title( "WORKING TABLE Background characteristics of respondents for the DV module"," ",
"Percent distribution of women who completed the domestic violence module by selected background",
"characteristics, Country 2011" )
stub( "Background characteristic" );

crosstab float(1) DV01 v013w11+v130w+v131w+v102w+v101w+v501w2+v218w+v741w3+v106w+v1 90w+tot1549 coldv01a+coldv01b+numwom
exclude(totals,specval,rowzero,colzero,percents)
{+US}
title( "Table DV.1 Experience of physical violence","",
"Percentage of women age 15-49 who have ever experienced physical violence since age 15 and",
"percentage who have experienced violence during the 12 months preceding the survey, by ",
"background characteristics, Country 2011" )
stub( "Background characteristic" );

PROC RECODE6_FF
preproc

numwom = 1;
numwom2 = 1;
numwom3 = 1;
ncurrm = 1;
emwoman = 1;
phsviol = 1;
twviol = 1;
total = 0;
tot1549 = 0;

unweight = ( sysparm()[1:1] = "U" ); { 1 - run unweighted tables }

postproc
{ constructs table to determine whether run is weighted/unweighted }

txxx(unweight,0) = sysdate( "dd" ); { day }
txxx(unweight,1) = sysdate( "mm" ); { month }
txxx(unweight,2) = sysdate( "yyyy" ); { year }

{ Table DVWT1 processing }
itot = tblrow( dvwt1 );
jmax1 = tblcol( dvwt1, imarr = 1 coldvwt1 = 1 );
jmax2 = tblcol( dvwt1, imarr = 2 coldvwt1 = 1 );
do i = 0 while i <= itot
dvwt1[i,jmax1] = dvwt1[i,jmax1+1] * 100 / dvwt1[itot,jmax1+1];
dvwt1[i,jmax2] = dvwt1[i,jmax2+1] * 100 / dvwt1[itot,jmax2+1];
enddo;

{ Table DV.1 processing }
jtot = tblcol( DV01 );
jmax = jtot - 1;
do j = 0 while j <= jmax
DV01[*,j] = DV01[*,j] * 100 / DV01[*,jtot];
enddo;
{ check unweighted N's }
Col2Dim( "dv01", dv01, 0, jtot-1, dv01u, 0 );

Previous Topic: Domestic violence and weights
Next Topic: Indicators of Physical and Emotional Violence
Goto Forum:
  


Current Time: Sun Jun 20 18:55:15 Coordinated Universal Time 2021