Creation of a realistic population for a microsimulation
Michelle A. Gosse, Mark S. Bebbington, and Jonathan C. Marshall
Institute of Fundamental Sciences, Massey University, Palmerston North
Population-based microsimulations using agents require external validity so that the simulation results approximate reality. The construction of a realistic population is particularly crucial for agent-based simulations. Agent assignment to households is a detail that tends to be glossed over in publications. This presentation will present an overview of the construction of a simulated New Zealand town to model disease transmission and spread from a food-borne illness. Four external datasets were used to construct the population, of which three were random-rounded for confidentiality reasons. As well as households, agents have been allocated to schools (children) and employers (employed agents). Most of the work has been implemented in R, using probabilistic methods disaggregate random-rounded data. Some useful tips for implementation in R will be covered.