The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Dataset use in Stata » Multi-level mixed effect Poisson regression in Stata
Multi-level mixed effect Poisson regression in Stata [message #23236] Wed, 04 August 2021 13:38 Go to next message
iqbalnowshad is currently offline  iqbalnowshad
Messages: 1
Registered: August 2021
Member
Dear User forum members,
I am using the individual DHS survey data (IR file) where subjects are nested within households, and households are nested within clusters. My outcome of interest is a count variable and I have several categorical predicts. I would like to fit a multi-level mixed effect Poisson regression model with robust variance to identify factors associated with the outcome in Stata (version 15.1).

However, I am having difficulty in writing the code in Stata. For survey data we usually use "SVY" command. However, Stata documentation (https://www.stata.com/manuals/memepoisson.pdf) shows below command for "mepoisson" multi-level mixed effect Poisson regression which is very confusing to me.

svyset psu, weight(wvar3) || ssu, weight(wvar2) || _n, weight(wvar1)
svy: mepoisson y x || psu: || ssu:

How do I incorporate weight (v005), primary sampling unit (v021) and strata (v022) variable in the above code? Also how to estimate values for ICC, PCV, AIC, and BIC after model fitting.

I was hoping someone could help me with this or if anyone has done similar analysis and shares the code that will be much appreciated.

Thank you,
Iqbal
Re: Multi-level mixed effect Poisson regression in Stata [message #23284 is a reply to message #23236] Mon, 16 August 2021 10:44 Go to previous message
Shireen-DHS is currently offline  Shireen-DHS
Messages: 136
Registered: August 2020
Location: USA
Senior Member
Hello,

Multilevel models require a weight for each level.

The generic code for instance for two levels would be the following:

svyset v001, weight(wt) strata(v023) , singleunit(centered) || _n, weight(wt2)

v001 is the psu and v023 is the strata if using DHS data. To check the strata for older surveys please refer to the Survey_strata.do file here: https://github.com/DHSProgram/DHS-Analysis-Code

We only supply one weight in the DHS data (v005). However, recently we have published a report to estimate the second level weight. Please refer to this report for more detail. It also contains examples and Stata code: https://www.dhsprogram.com/publications/publication-mr27-met hodological-reports.cfm

Good luck with your research.

Best,
Shireen Assaf
The DHS Program
Previous Topic: Community level Poverty
Next Topic: How to carry out Individual Meta-analysis of DH surveys in sub-Saharan Africa using Stata
Goto Forum:
  


Current Time: Wed Feb 1 07:46:26 Coordinated Universal Time 2023