Multi-level regressions with survey data

Multi-level regressions with survey data, a topic that is a far from trivial. Survey data can be tricky to analyse because it has many discontinuous variables, and missing data, which introduces quite a bit of uncertainty in the results as well us increasing heterogeneity. In general, social country-scale surveys hide so much noise that make everything significant while prodicing barely measurable marginal effects, which are not convinsing, as a result. To demonstrate how to report marginal effects from large scale survey data, I post an R code, where I use World Values Survey data from wave 6 for 50 countries to explore relationship between respondents' self-reported protest participation and online media consumption. I run multi-level logistic regression and simulations to estimate marginal effects.