bensoltoff
commented
5 years ago - I'm a bit confused as to your measure of democratic backsliding. You state that countries will be classified as democratic backsliders if they are "partially democratic". Does that mean they were once full democratic and are no longer that highly rated? If not, couldn't a country transition from autocratic to partially democratic and stall out there? If they were never fully democratic, how could they have backslid?
- Why just a one-year lag in the predictive model? Provide justification for that level
- Your "sample decision trees". Are these separately estimated from the RF models, or one of the trees from the forests?
- RF models don't drop missing value observations. Remember that it has an alternative treatment method for them to leverage their non-missing values. You can impute the missing values, but it isn't necessary like what you would have to do in an OLS or logistic regression model