We are trying to fit a multi-level logistic regression and we work with public opinion data. Our dependent variable is "interest in politics", a dummy variable. We have individual and group-level (country) covariates. Hence, we work with group level predictors. Actually, our theory argues that, the interaction between income inequality and populist president (i.e. presidente*gini) positively affects individuals? interest in politics (we attach the core of our argument). We want to know if this is the correct way to write the model with R. We have not seen examples with more than one predictor at the group level. Also, group level is mostly included to improve the fit of the model. lapop10.out.2<-lmer(pol11~Q1+Q2+Ur+Q10+Ed+presidente*gini+I(pib/50)+elecc+(1+Q1+Q2+Ur+Q10+Ed|Pais), family=binomial(link=logit), data=lapop) pol11: interest in politics (dummy variable at the individual level) Q1: sex (individual level) Q2: age (individual level) ed: education in years (individual level) presidente: dummy variable that accounts for the presence (absence) of populist president (country level) gini: gini index (country level) pib: pib per capita (country level) elecc: dummy variable to control whether 2010 was an electoral year (country level) Also, we work with 18 Latin American countries. In 2010, in each country a Lapop (Latin American Public Opinion Project) survey was conducted and identical questions were asked. In this sense, we also want to know whether working with all these country level variables (6) is problematic or not. If we include more survey years (2006, 2008), how should we specify the model with R (i.e. a three-level model: individual-year,country)? Thank you Fernando Rosenblatt Rafael Pi?eiro PUC-Chile -- View this message in context: http://r.789695.n4.nabble.com/Fitting-a-multi-level-logistic-model-tp3676362p3676362.html Sent from the R help mailing list archive at Nabble.com.