WeightedCluster Library Manual: A practical guide to creating typologies of trajectories in the social sciences with R

TitreWeightedCluster Library Manual: A practical guide to creating typologies of trajectories in the social sciences with R
Type de publicationJournal Article
Year of Publication2013
AuteursStuder, M
JournalLIVES Working Papers
Volume2013
Ticket24
Pagination1-32
PublisherNCCR LIVES
Place PublishedLausanne
Type of ArticleTechnical report
ISSN2296-1658
Mots-clésanalysis of sequences, cluster, cluster quality measure, distance, life course, optimal matching, R, trajectory, typology, weighting
Résumé

This manual has a twofold aim: to present the WeightedCluster library and offer a step-by-step guide to creating typologies of sequences for the social sciences. In particular, this library makes it possible to represent graphically the results of a hierarchical cluster analysis, to group identical sequences in order to analyse a larger number of sequences, to compute a set of measures of partition quality and also an optimized PAM (Partitioning Around Medoids) algorithm taking account of weightings. The library also offers procedures to facilitate the choice of a particular clustering solution and to choose the optimal number of groups.

In addition to the methods, we also discuss the building of typologies of sequences in the social sciences and the assumptions underlying this operation. In particular we clarify the place that should be given to the creation of typologies in the analysis of sequences. We thus show that these methods offer an important descriptive point of view on sequences by bringing to light recurrent patterns. However, they should not be used in a confirmatory analysis, since they can point to misleading conclusions.

DOI10.12682/lives.2296-1658.2013.24