The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Countries » Nigeria » Estimating and Matching Under5 Mortality by cluster units
Estimating and Matching Under5 Mortality by cluster units [message #8516] Sat, 07 November 2015 20:11 Go to next message
MildredA
Messages: 3
Registered: November 2015
Location: United Kingdom
Member
Hello Everyone,

Apologies for asking what could be basic questions to many. This is because I am new to DHS data and statistics as a whole.

I need help with estimating under5 mortality rates for all survey clusters for Nigerian DHS since 2003. I have studied the guide to dhs statistics and the recode manual and could not understand it. I would be very grateful if anyone could help me identify the variables and the steps/equation I need to compute U5 mortality rates in very simple terms in SPSS.

Also, I am not sure which recode file (birth, household, individual or children) to use to generate the U5 mortality rates. How do I aggregate categorical variables such as ethnicity and age groups by clusters such that, each category becomes a new variable expressed in percentages for all cluster points.


Since, I will be merging data from 2003-2013, I will like to know whether the same sample clusters are used to collect information for the the three DHS years in Nigeria.

Please help me out, my dissertation is hanging on this.

Kind regards

Mildred Ajebon
Re: Estimating and Matching Under5 Mortality by cluster units [message #8523 is a reply to message #8516] Sun, 08 November 2015 16:56 Go to previous messageGo to next message
Reduced-For(u)m
Messages: 290
Registered: March 2013
Senior Member

Mildred,

There is a lot to answer here, but a question first:

Do you want to estimate "cluster" (meaning Primary Sampling Unit) averages of child mortality? That will be impossible given cluster-specific sample sizes. In general, you need thousands of children to accurately estimate that rate, and there are only a handful of children per cluster (more if you go way back in time, but then you are getting cluster- and time- differences).

Also - DHS sampling units generally changes from survey to survey, so you can't usually compare them over time.

If I'm wrong about either of these assumptions regarding what you are doing, let me know. If not, I don't think this project is feasible.
Re: Estimating and Matching Under5 Mortality by cluster units [message #8524 is a reply to message #8523] Sun, 08 November 2015 18:25 Go to previous messageGo to next message
MildredA
Messages: 3
Registered: November 2015
Location: United Kingdom
Member
Thanks a million for your response.

Firstly, the most important huddle for me to cross at this point is to generate proportion of deaths (under5, maternal and if possible, adult mortality)for different geographical units within Nigeria. That is, the risk of under5, maternal and adult mortality in area clusters/LGAs. Technically, I want to be able to generate a comparative measure of mortality for geographical areas within Nigeria and perhaps, other DHS countries. This will help me understand within-country variation in health.

I would be grateful if anyone could help suggest which recode file to use, what variables to include and how to compute these in SPSS. Merging data over time was something is not the central theme of my work. I could do away with that. It was something extra I thought I could do.


Secondly, if cluster points are not maintained across surveys, what is the best way to compare risk of dying across geographical areas for the different survey years? Is it feasible to use Local Government Area (LGA)boundaries to map the distribution of under5 mortality?. I am a geographer, it is important for me to be able to map results for the smallest area unit possible for visualization.

Regards









[Updated on: Sun, 08 November 2015 18:56]

Report message to a moderator

Re: Estimating and Matching Under5 Mortality by cluster units [message #8606 is a reply to message #8524] Fri, 20 November 2015 16:12 Go to previous message
Trevor-DHS is currently offline  Trevor-DHS
Messages: 626
Registered: January 2013
Senior Member
You cannot compute the mortality rates at the level of LGAs. The sample was not designed to be representative at these levels and the sample size is too small to give you valid results at this level. The mortality estimates can only be validly produced for the disaggregations presented in the report.
Previous Topic: Generating and Matching Under5 Mortality by cluster units
Next Topic: 2013 Nigeria DHS analysis on IPTp uptake
Goto Forum:
  


Current Time: Tue Nov 13 16:09:35 Eastern Standard Time 2018