Searching for typical life trajectories applied to childbirth histories
Alexis Gabadinho, Gilbert Ritschard · 2013
Abstract. We address, in this chapter, the identification of typical patterns that best characterize a set of sequences. More specifically, we focus on data-driven methods that search for the typical patterns among the observed sequences. In life course studies, such typical sequences serve, for instance, to describe ideal-type life trajectories; i.e., the common way(s) of organizing our life. We propose an heuristic for extracting an as small as possible subset of patterns that covers a given percentage of all sequences. The method is based on the concept of neighborhood, the coverage being the number of representative sequences in the neighborhood of the retained representative patterns. We illustrate the scope of the method by applying it to the study of childbirth histories of Swiss women. The representative sets obtained for six successive birth cohorts clearly exhibit that while patterns with three or four childbirths were common for women of older cohorts, it is no longer the case for younger cohorts. At the same time, representative plots permit to identify the typical timing of the successive childbirths.