Dear all,

I am currently working with a cluster data collected from 3229 respondents in 40 counties, half of which are policy pilot sites. The dependent variable is categorical data from 1-5 regarding individuals' attitudes.

I am trying to fit a 3 level mixed-effect model by using the demand of meologit in Stata 15.

The first level: SES + explanatory variable about individuals
The second level: 40 country id
The third level: pilot (0 or 1)
code: meologit urbani A1b A0 marriage i.educat4 occu_prof hhwealth hukou_cat32 hukou_cat33 hukou_out_town_impute1 [pw=weight_temp_use] || pilot1t: || gbcode: , nolog

I am thinking the 3229 individuals are nested with country, and then the 40 counties nest with the pilot, but not sure for the following questions:

1. Is the binary variable (pilot) can be used as the third level control? Especially when considering the group size issue.
(now the third level data is divided into two groups: 20 pilot country and 20 non-pilot country)
2. The results as following show that the variance of pilot is quite large, and what can we tell?
3. Also, there did not show the test result in front of the table...how should I modify my model or code accordingly?

Thank you for your reply if you are familiar with this model.

I really appreciated your help.


Array