Re: st: two-level dataset, binary outcome, 4 level-2 variables

[Amir Sariaslan <amir.sariaslan@gmail.com>]

Dear Rik,

It is arguable whether or not you have nested data given that you
don't have any level-1 covariates. I would not attempt to model your
dataset using multilevel modeling. One alternative route is to look at
counts of sick individuals in the cities. I'm thinking of a
single-level Poisson regression model (-poisson- or any related model
that accounts for dispersion) where you adjust for your four
covariates and an additional one that measures the population size of
the cities.

Sincerely,
Amir


On Tue, Jan 4, 2011 at 5:27 PM,  <Roderik.Viergever@lshtm.ac.uk> wrote:
> Dear stata users,
>
> I am a PhD student from London, and would appreciate your advice on a two-level dataset. We have a binary outcome (sick/healthy) for individuals in several cities. Every city has a different number of individuals. We have four variables, they are all on the city-level, so we have no individual level variables. We wish to test if the variables are predictive of the outcome sick/healthy.
>
> Question 1: Simple logistic regression does not seem to be able to take into account the different numbers of individuals per city, so we currently are of the opinion that multilevel logistic regression is the appropriate analysis for us. But is this really necessary, given that we do not have any individual level variables?
>
> Question 2: IF multilevel logistic regression is the way to go, would this be the correct command for STATA?
>
> xtmelogit  binary_outcome ||  city_list:  Var_1 Var_2 Var_3 Var_4
>
> Many thanks in advance for your advice,
> Rik
>
>
> *
> *   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/