I need to run a regression analysis using a data set on incidents of war. Data are collected on a country-year format, and go back hundreds of years.
My question is: some countries have gone to war only once, whereas others fought 10 or more. Under the circumstances, it feels wrong to run the analysis without any weights, as some countries will be over-represented.
Can appropriate weighting be a solution to this problem? If so, how should I apply the weights? Is it okay for a country that fought 10 times (and thus has 10 rows in the data set) to have a corresponding weight of 1/10 for each row? Accordingly, a country that fought 7 times would have a weight for 1/7 for each row, and so on. Is this the right approach?
