The DHS Program User Forum
Discussions regarding The DHS Program data and results
Today's Messages (on)  | Unanswered Messages (off)

Forum: General
 Topic: Calculating CI for median survival time
Re: Calculating CI for median survival time [message #18222 is a reply to message #18043] Thu, 17 October 2019 10:26
Liz-DHS
Messages: 1401
Registered: February 2013
Senior Member
A response from Dr. Ruilin Ren, Senior Sampler:
Quote:
Thanks for the post. If you use any of the major statistical software such as SAS, SPSS and STATA, you should be able to calculate the sampling error and the confidence interval of median using the complex survey module. In DHS, we use SAS. If you use SAS, you can use the codes:

PROC SURVEYMEANS median DATA=your data;
VAR XXX (name of your edited survival time variable);
WEIGHT YYY (name of your weight variable);
STRATA ZZZ (the stratification variable);
CLUSTER V001;
RUN;

It is worth to note that SAS calculates a true interpolated median, while SPSS and STATA calculate the median value of the variable, not an interpolated median.
Hope this helps.
Ruilin Ren

Forum: Other countries
 Topic: Postnatal care - Turkey 2013
Postnatal care - Turkey 2013 [message #18221] Thu, 17 October 2019 09:17
ICEH is currently offline  ICEH
Messages: 11
Registered: May 2017
Member
Hello,

My group and I are trying to replicate the results published in Table 10.7 of Turkey 2013 Report.
We are using variable m51, restricting to values 100 to 148, 201 and 202 (postnatal care in the first two days of birth).
However, our estimate has a huge diference in relation to the published one.

We tryied to estimate it considering the last birth in the five years preceding the survey, to replicate the published estimative, but the result was 52.8% (N=2,603).

Using the standard definition, considering women age 15-49 giving birth in the last two years preceding the survey, result was 54.8% (N=1,227).

What can we be considering different from Turkey 2013 report?

Thanks in advance,


ICEH
 Topic: Child development - Benin 2017
Child development - Benin 2017 [message #18223] Thu, 17 October 2019 10:49
ICEH is currently offline  ICEH
Messages: 11
Registered: May 2017
Member
Dear all,

I want to estimate child development indicators using DHS 2017 Benin, however, my denominator (variable b19 = 36 to 59) is 4415 children, much higher than the published in table 18.12.
I tried to restrict the denominator to only one child age 36-59 months in households with more than one child at this age, but the number still higher than 4000.
I also tried to reestimate the current age of the child in months, and matches with b19 - so it seems that it is no problem there.

Is there another restriction?


Thanks in advance,

ICEH
 Topic: Senegal 2012 regions
Senegal 2012 regions [message #18224] Thu, 17 October 2019 10:52
dhsLearner is currently offline  dhsLearner
Messages: 3
Registered: October 2019
Member
I am looking at the 2012 Senegal data and trying to summarize various indicator variables by the 14 regions. I see that hv023 (Stratification used in sample design), hv102, hv024 list these 14 regions. However, the 2012 report summarizes data by 4 regions (centre, nord, ouest, and sud), and the spatial repository also only includes data summarized to these four features. Would it be appropriate for me to summarize variables by the 14 regions? Thank you for your help.
Forum: Nutrition and Anthropometry
 Topic: Can't get exact match between one regional calculation and reported stunting value
Can't get exact match between one regional calculation and reported stunting value [message #18225] Thu, 17 October 2019 12:38
dhsLearner is currently offline  dhsLearner
Messages: 3
Registered: October 2019
Member
I am trying to calculate childhood (under 5) stunting rates by region in Senegal in 2012. For one region, I am getting a different result than what it is reported in the Senegal sumary report. Here is my R code:

#read in data
PR2012<-fread("..\\dhs\\dataForR\\continuousSurvey2012\\SNPR6DFL.csv ")

# Create a binary variable for children stunted
PR2012$stuntedUnderFive<-ifelse(PR2012$hc70< -200, 1, 0)
# Assign NA when z score is >9900 sd (NA) #9998 are don't know responses
PR2012$stuntedUnderFive<-ifelse(PR2012$hc70>9900|
PR2012$hc1> 59, NA, PR2012$stuntedUnderFive)

# Create weight variable
PR2012$wt<-PR2012$hv005/1000000

# Tabluate stunted by region
PR2012%>%
group_by(shzone)%>% #shzone is the four regions (north, east, south west)
summarize(meanStunting = weighted.mean(stuntedUnderFive,wt,na.rm = T))

Results:

1 centre 0.190
2 nord 0.168
3 ouest 0.145
4 sud 0.266

I calculate West (ouest) as 14.5%, but it is reported as 14.7% in the summary (https://dhsprogram.com/pubs/pdf/FR288/FR288eng.pdf).

Can anyone help me figure out what I am doing incorrectly?
Re: Can't get exact match between one regional calculation and reported stunting value [message #18226 is a reply to message #18225] Thu, 17 October 2019 14:16
Trevor-DHS is currently offline  Trevor-DHS
Messages: 670
Registered: January 2013
Senior Member
Can you check your denominator? I suspect the problem is in the selection of children. You should be restricting to de facto children (hv103 == 1). Check the denominator first and then if you are still finding a problem, let us know.
Re: Can't get exact match between one regional calculation and reported stunting value [message #18227 is a reply to message #18226] Thu, 17 October 2019 14:25
dhsLearner is currently offline  dhsLearner
Messages: 3
Registered: October 2019
Member
Thank you. This was the problem. Wow there are so many fine print notes and variables that go along with them. Definitely need to keep my eyes out for filters like this.

For future reference to other people looking at this forum post, I added hv103 (did child stay in house night before) to my code as follows:

PR2012$stuntedUnderFive<-ifelse(PR2012$hc70>9900|
PR2012$hc1> 59 | PR2012$hv103 =="no", NA, PR2012$stuntedUnderFive)
Forum: Dataset use in Stata
 Topic: 2017 Ghana Maternal Health Survey: Restricting Sample to Individual Pregnant Women (5-year Period)
2017 Ghana Maternal Health Survey: Restricting Sample to Individual Pregnant Women (5-year Period) [message #18219] Thu, 17 October 2019 00:05
ebanchani is currently offline  ebanchani
Messages: 9
Registered: January 2019
Member
I'm using the 2017 Ghana Maternal Health Survey, how can I limit my sample to the number of individual pregnancies within the five period preceding the survey. My focus of analysis is individual women who reported pregnancies within the five-year period and not on the pregnancy history.
Thank you.
 Topic: missing variables in 2017 GMHS
Re: missing variables in 2017 GMHS [message #18228 is a reply to message #17736] Thu, 17 October 2019 19:55
ebanchani is currently offline  ebanchani
Messages: 9
Registered: January 2019
Member
Hello abiney,
I'm also working on the 2017 GMHS and I want to asked some few questions if you can help me out:
How can I restrict the sample of women who reported a pregnancy in the five (5) year preceding the survey?
I will appreciate your timely response.
Best.

Emmanuel



Current Time: Thu Oct 17 22:19:02 Eastern Daylight Time 2019