Bootstrapping approaches for validation of a spatial microsimulation model in Australia

Validation and creation of the statistical reliability of small area estimates by microsimulation modelling is challenging due to the fact that the estimated data are typically unavailable from another source for the small areas. The literature suggest few methods including aggregating the small area estimates up to larger levels where reliable data are available to compare the results with the attempt of adopting alternative methods to determine the accuracy of the model (e.g. Tanton et al., 2014; Timmins and Edwards, 2016; Rahman and Harding, 2016; Whitworth et al., 2016). Bootstrapping approaches (Efron and Tibshirani, 1993; Steyerberg et al., 2003) can be another choice for validation of the estimates for small areas with smaller populations. Although bootstrapping is based on re-sampling technique, and widely used as methods of validating statistical prediction models (Gerds, et al., 2008; Kopec et al., 2010), bootstrapping approaches have not, to our knowledge, been applied to validate spatial microsimulation models. This research aims to introduce bootstrapping methodologies in calculating statistical reliability measures for small area estimates from a spatial microsimulation model in Australia. An empirical application demonstrates that the bootstrapping approaches seem most meaningful systems for validation of the spatial microsimulation model for those areas with smaller populations. Some key advantages of such methods are also explored.