Nieves uses random forest machine learning to predict population density. This is done by first taking the datasets of various covariates as well estimated population density distributed across an administrative layer. These datasets are combined to form a ‘bootstrap sample’. Then, decision trees involving the combination of datasets are ‘grown’ using machine learning, which when combined, will create a simulated population density distribution that was more accurate then the original census. Finally, random forest takes a small ‘out of bag’ sample from the original ‘bootstrap sample’, which is compared to the final model to estimate a margin of error. A dasymmetric allocation is the process of distributing parts of the total population density in a non symmetrical way. The article concluded that the most important covariates to estimate population density were topographical features.