Replication Data for: Measuring Arms: Introducing the Global Military Spending Dataset

Agreement between observed country-year-variable values (orange boxes) and the posterior predicted point estimates for which the observed value is observed (light grey boxes). Dark grey boxes show the distribution for the full range of the posterior predicted point estimates (including estimates for which the original value is missing). Across all variables, these estimated values have a lower median value due to a bias in the missingness in the original data. We have more missing observations in earlier years, for which population levels are lower than in later years in the series. Conveniently for users, the posterior predicted values are estimated using the original unit-of-measurement. This means that the visual discrepancy is not an empirical discrepancy. It is simply the difference in the unit-of-measurement used for the observed dataset values (e.g., thousands of dollars units).