Abstract
We develop a new algorithm for population synthesis that fuses remote-sensing data with partial and sparse demographic surveys. The algorithm addresses non-binding constraints and complex sampling designs by translating population synthesis into a computationally efficient procedure for constrained network growth. As a case, we synthesize the rural population of Afghanistan, validate the algorithm with in-sample and out-of-sample tests, examine the variability of algorithm outputs over k-nearest neighbor manifolds, and show the responsiveness of our algorithm to additional data as a constraint on marginal population counts.
Acknowledgments
We thank three anonymous reviewers for constructive comments and suggestions, and Hector Maletta for valuable survey data on Afghan agriculture. The Office of Naval Research (ONR) grant N00014-08-1-0378 partially supported this research. Views expressed herein are ours, not of George Mason University or the ONR.