Little problem on checking my null model and ICC for logistic model with multilevel

[Anita Desmarini]

I am using stata 13 to run my data. The data is consisted of 45.607 observations. It is binary logistic with multilevel. When I was running my null model (which is ONLY include of respon variable and variable for level 2, WITHOUT any predictor), It yielded an error message as following :

initial values not feasible
r(1400);

----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

when I tried another respon variable with 9297 observations (It's also binary logistic with 2 levels). I just checked the null model, then the result could appear.

----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Both were collected from the same resource. Both were binary logistic with the same level (I used PROVINCE as variable for level 2). But why the result for the first case didn't appear ? Does anyone know. I will really appreciate every answer that you give

----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

btw, here's the detail :

. melogit Ybaru_tanpaMissing || V024:

Fitting fixed-effects model:

Iteration 0: log likelihood = -28790.719
Iteration 1: log likelihood = -28719.134
Iteration 2: log likelihood = -28719.093
Iteration 3: log likelihood = -28719.093

Refining starting values:

Grid node 0: log likelihood = -28543.163

Fitting full model:

initial values not feasible
r(1400);

[Carole J. Wilson]

This question is not related to Mata (Stata's matrix language). You should re-post to the general Stata forum: http://www.statalist.org/forums/foru...ussion/general