Hello everyone,

My name is David. I am new here. I need your help to solve the command issues that I am currently facing in Stata 14.

Specifically, I am having problems executed the "menbreg" command to fit the multilevel negative binomial regression to model "the number of children dead" (per woman) with exposure being years since age at first birth (i.e. mothers age minus age at first birth). I had to settle for "menbreg" because I couldn't get help on how to run multilevel zero-inflated negative binomial regression in Stata.

I am analyzing secondary data of 29,631 observations, 8 individual-level variables, 7 household-level variables and 10 community-level variables. The second level is the household ID (v002) while the third level is the cluster ID (v001.

When I run the syntax below, I get this error notification: initial values not feasible r(1400);


Code:
****Individual level
global individual v012 v212 i.mari i.v133 unemp i.religion i.ethnicity contra

****Household*****
global household v201 i.num_mal i.mosq i.elect i.wealth i.pipe i.flush 

****Community
global community rur i.comm_diversity i.region i.comm_uneduc i.comm_poverty i.comm_nonusecont i.comm_pipe i.comm_elec i.comm_toilet prob_dist


****Model 1 (Individual only)****
menbreg dead_children $individual, exposure(exp_age_at_first_birth) || v002: || v001:, irr

****Model 2 (Household)****
menbreg dead_children $individual $household, exposure(exp_age_at_first_birth) || v002: || v001:, irr

****Model 3 (Community)****
menbreg dead_children $individual $household $community, exposure(exp_age_at_first_birth) || v002: || v001:, irr

The image below is what is the outcome of the first model and also the same when I run the third model.

Array


I would be glad to get help and direction given that this is my first time of employing such model which isn't out of place considering the over-dispersion in my outcome variable (number of dead children) and the multilevel design of the survey data.

Thank you.