Multilevel Poisson

[Khawar Afreen]

I am working on multilevel Poisson. I faced a problem with the svy command when I ran mepoisson. As I am working with the DHS dataset. I have calculated level-1 and level-2 weights according to the DHS report 27. Thereafter, I ran the mepisson command. But mepisson does not support svy. I have also followed the Stata manual for multilevel Poisson modelling to create the command mentioned below. If anyone could, please help me in this matter. Is my mepoison command correct?


* Create a constant cluster-level weight
bysort v001: egen wt2_05_constant = mean(wt2_05)

mepoisson children_born i.dv_any_violence i.region v012 i.education i.residence ///
lom_prof_community ///
[pweight=wt1_05] || v001:, pweight(wt2_05_constant) irr vce(robust)

[Rich Goldstein]

if you go to

help svy_estimation

you will see that -mepoisson- is supported

maybe you are using an old version of Stata? if so, you need to tell us

[Khawar Afreen]

Dear Rich,

Thank you for your quick response. I am using Stata 17.

I don't understand why my command is not working. The correct way that DHS report 27 is guided is

* Set the survey design using the alpha 0.5 weights
svyset v001, weight(wt2_05) strata(v023) , singleunit(centered) || _n, weight(wt1_05)

svy: mepoisson children_born i.dv_any_violence i.residence v012 i.education ///
lom_prof_community || v001:


But after running this command, I received the following error:

* Set the survey design using the alpha 0.5 weights
svyset v001, weight(wt2_05) strata(v023) , singleunit(centered) || _n, weight(wt1_05)
note: stage 1 is sampled with replacement; further stages will be ignored for variance estimation.

Sampling weights: <none>
VCE: linearized
Single unit: centered
Strata 1: v023
Sampling unit 1: v001
FPC 1: <zero>
Weight 1: wt2_05
Strata 2: <one>
Sampling unit 2: <observations>
FPC 2: <zero>
Weight 2: wt1_05

svy: mepoisson children_born i.dv_any_violence i.residence v012 i.education ///
> lom_prof_community || v001:
(running mepoisson on estimation sample)

weights in variable wt2_05 not constant within groups defined by: v001
an error occurred when svy executed mepoisson
r(198);


Please guide me to solve this issue

[Rich Goldstein]

sorry, but I don't understand what your data are like at all; I am also having trouble understanding what you are trying to do; please read and follow the advice in the FAQ

[Khawar Afreen]

Dear Rich,

I am using the DHS dataset. v001 is PSU; weight (wt1_05) is level one weight; and weight (wt1_05) is level two weight, which are calculated according to the guidelines of the DHS report. I am applying a multilevel mixed-effect Poisson.
Below is the guideline for the multilevel Poisson Stata manual with the svy command, which works for level 3 weights. There is no code for multilevel Poisson with svy for two-level weights.

Two-level random-intercept model from a two-stage sampling design with PSUs identified by psu
using PSU-level and observation-level sampling weights wvar2 and wvar1, respectively
mepoisson y x [pweight=wvar1] || psu:, pweight(wvar2)

Add secondary sampling stage with units identified by ssu having weights wvar2 and PSU-level
weights wvar3 for a three-level random-intercept model
mepoisson y x [pw=wvar1] || psu:, pw(wvar3) || ssu:, pw(wvar2)

Same as above, but svyset data first
svyset psu, weight(wvar3) || ssu, weight(wvar2) || _n, weight(wvar1)
svy: mepoisson y x || psu: || ssu: