Thanks so much for your response, Steve. I studied those very links and
finally managed to come really close to the DHS numbers. They're not
exactly the same yet, but I suspect that the way Stata handles weights (as
opposed to the program DHS used) might have something to do
with the minor differences . I am now within +1/-1 of their rates. At
least I have defensible numbers now.
Thanks again for your help.
Satvika
> -
> [My last posts have not gone through, so I am sending this one from a new
> wemail Statalist address. I apologize for the duplication if earlier posts
> also appear.--Steve]
>
>
> Satvika-
>
> Why do your results differ? Perhaps because you did not duplicate the
> definitions or calculations that DHS uses.
>
> See:
> http://www.measuredhs.com/help/Datasets/Infant_and_Child_Mortality.htm
> especially
> http://www.measuredhs.com/help/Datasets/Mortality_Rate_Calculation.htm
>
> To define neonatal mortality, for example, Measure DHS does not use 28
> days, but 30 days. Also it does not use seven day intervals. Instead, it
> computes numerator and denominator for the period 0-30 days. It then
> subtracts from the denominator 1/2 for each child born <30 days prior to
> the five years which precede the survey. You must define these births as
> censored, and use -ltable-, not -stlist-. To allow children still alive
> after 30 days to contribute fully to the denominator, you must have age
> intervals after 30 days in your analysis.
>
> Censoring patterns for ages >30 days are more complicated, as the second
> web page shows. Your code might also have excluded some observations
> unnecessarily, because while Measure DHS imputes missing values, the
> imputed values may not always be exact dates. Perhaps the study code book
> will enlighten you.
>
> On the web pages referenced, there is no mention of weighting, but I would
> guess that the counts are weighted.
>
> Good luck!
>
> -Steve
>
>
> On Mar 9, 2009, at 6:11 AM, Satvika Chalasani wrote:
>
>
> Hello Statalisters,
>
> I am trying to reproduce the neonatal/infant/under-5 mortality rates shown
> in the Demographic and Health Survey reports, particularly for the Indian
> surveys. Now DHS does provide SPSS and SAS modules to calculate mortality
> rates their way but not a Stata module. I am using sts list to calculate
> failure functions, and in two surveys I come close to the published rates
> but in one I don't. ltable seems even less accurate. Does anyone have a
> routine that they would be willing to share? Or any thoughts?
>
> My code:
>
> gen ageatexit=.
> replace ageatexit=(interviewdate-dateofbirth) if childalive==1
> replace ageatexit=(dateofdeath-dateofbirth) if childdied==1
>
> preserve
> stset ageatexit [pw=weight], failure(neonatal) id(caseid) exit(t 28)
> stsplit dur, every (7)
> gen neonatal_d=0 if _d~=.
> replace neonatal_d=1 if _d==1
> sts list, failure
> restore
>
> Thank you!
> Satvika Chalasani
>
> Graduate student
> Penn State University
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
