No announcement yet.
  • Filter
  • Time
  • Show
Clear All
new posts

  • Fixed effects vs random effect (NBM) and stata command

    I am investigating the effect of Human Development Index (country level Independant variable) on Negative experienced discrimination (dependant variable) in 807 participants with schizophrenia who completed a discrimination survey in 29 countries. My dependant variable is a count variable - total subscore of negative experienced discrimination is calculated by totalling up each time a participant endorsed negative discrimination on each item of the survey. I have concluded that i will model this using a negative binomial regression as there is over-dispersion therefore NBR is better than a poisson model. As the data is clustered at the country level, I also think i need to use a random intercept model which explicitly models the dependence of individual observations within clusters by using a country specific random intercept. I will then be adding other sociodemographic covariates into the model e.g age, gender, years of education, current employment, years since first contact with health services, ever had compulsory treatment, do you know your diagnosis and whether they agree with their diagnosis. Are these covariates all fixed effects? Can anyone help with the stata command that i need to build the model. So far i have this for the univariate model "xtnbreg negdisc HDI, re i(country) base" not sure what the command would look like for the multivariable model