Weights with gsem

[Amanda Wyant]

Hello,
I am trying to fit a multi-level multinominal logit model using gsem. I have weights both levels of my data (Level 1 -individuals and Level 2 - countries). I am having trouble finding the correct placement of the weights within the code for gsem. What should the code look like to incorporate two levels of weights within gsem? I'm using the current version of Stata (14.2).

Thank you!

[eric_a_booth]

Consider svyset-ing your data and using this approach (assuming you have stata 14): http://www.stata.com/new-in-stata/mu...s-survey-data/

[Amanda Wyant]

I'm not sure if I'm doing it correctly because my null model will not run. It just continues to run.

My code is:

svyset countryyr, weight(L2weight4),||interviewnum, weight(L1weight4)

svy: gsem (1b.revfood 2.revfood 3.revfood 4.revfood <- M1[countryyr]), mlogit

My data is people (interviewyear) nested within countries (countryyr). I also have weights at both levels. Am I setting it up incorrectly?

My output for the svyset:

. svyset countryyr, weight(L2weight4),||interviewnum, weight(L1weight4)
Note: Stage 1 is sampled with replacement; further stages will be ignored for variance estimation.

pweight: <none>
VCE: linearized
Single unit: missing
Strata 1: <one>
SU 1: countryyr
FPC 1: <zero>
Weight 1: L2weight4
Strata 2: <one>
SU 2: interviewnum
FPC 2: <zero>
Weight 2: L1weight4

[Jon Phillips]

Hello, I am wondering if you ever figured how to do get your model to run? I am also trying to figure out how to do a multilevel multinomial regression with sampling weights.

[Amanda Wyant]

Hi sorry for the long delay. I ended up making my dependent variable dichotomous and using melogit instead. If you figured it out, I'd love to hear about it!

[Mitch Lingo]

Another year later and here I was hoping someone would have figured this out by now.