ABSTRACT
The selection of linking variables should not be arbitrarily preordained by preconceptions. The decision of which constraints to use to allocate individuals to zones should be context dependent. If the research is on social exclusion, for example, many variables could potentially be of interest: car ownership, house tenancy, age, gender and religion could all affect the dependent variable. Often spatial microsimulation is presented in a way that suggests the data arrived in a near perfect state, ready to be inserted directly into the model. The ipfp reweighting strategy is concise, generalisable and computationally efficient. The weights produced by ipfp and mipfp packages are the same. The spatial microdata is thus multilevel data, operating at one level on the level of individuals and at another on the level of zones. To generate summary statistics about the individuals in each zone, functions must be run on the data, one zone at a time. aggregate provides one way of doing this.
