Extract

← 914 | 915 →

2.  Appendix 2: Higher-rank order agglomeration

Given the practical considerations discussed in Ch. 2.2, the search for regularities and similarities between the regions that would allow us to group them into units of a higher order – and, in turn, to decrease the number of objects included in each presentation of results – became a vital issue to be addressed. This section provides detailed information on how this goal was eventually achieved.

Our point of departure was the presence of twelve regions which came into existence by grouping parishes and estates on the basis of their administrative affiliation. We also made some intuitive proposals for agglomerating these territories into objects of a higher order based on some broad non-demographic characteristics (Ch. 2.2). The most pertinent issue at that stage was how best to assess the extent to which such an ‘intuitive grouping’ reflected broader spatial congruencies in family demography, and to what extent it risked pooling together objects which, although similar in terms of their administrative affiliation or their socioeconomic or ecological features, may have varied substantially in terms of their basic parameters of family organization.

There are several ways to make such an assessment (cf. Szołtysek and Biskup 2008a). We could, for example, imagine that each of our regions represents a sample of its constitutive parishes (or estates), and that for each parish in the collection we should estimate the average proportion of multiple-family households (or any other type of...