Random property allocation: A novel geographic imputation procedure based on a complete geocoded address file

Allocating an incomplete address to randomly selected property coordinates within a locality, known as random property allocation, has many advantages over other geoimputation techniques. We compared the performance of random property allocation to four other methods under various conditions using a simulation approach. All methods performed well for large spatial units, but random property allocation was the least prone to bias and error under volatile scenarios with small units and low prevalence. Both its coordinate based approach as well as the random process of assignment contribute to its increased accuracy and reduced bias in many scenarios. Hence it is preferable to fixed or areal geoimputation for many epidemiological and surveillance applications.