Harpoon or Bait? A Comparison of Various Metrics in Fishing for Sequence Patterns

Abstract

The use of sequence analysis in the social sciences has significantly increased during the last decade or two. Sequence analysis explores and describes trajectories and “fishes for patterns” (Abbott, 2000). Many dissimilarity metrics exist in various domains (bioinformatics, data mining, etc.); therefore a crucial and pervasive issue in papers using sequence analysis is robustness. To what extent do the various techniques lead to consistent and converging results? What kinds of patterns are more easily fished out by each of the metrics? Here we propose a systematic comparison of about ten metrics that have been used in the social science literature, based on the examination of dissimilarity matrices computed from a simulated sequence data set including various patterns that sociologists can try to identify. This should help scholars in picking the method best suited to their data design and inquiry objectives.

Do Different Methods for Modeling Age-Graded Trajectories Yield Consistent and Valid Results?

“…Robette and Bry (2012) recently compared results obtained using different forms of sequence analysis (of which OM is the most widely used), and identified variants that converged on consistent results in analyses of a simulated dataset. Barban and Billari (2012) analyzed both real and simulated datasets to compare results obtained from OM and LCA.…”

Comparing methods of classifying life courses: sequence analysis and latent class analysis

“…These metrics measure the distance between sequences by counting the number of (weighted) common subsequences. For a detailed review of distance metrics for SA, we refer to Robette and Bry (2012), Studer (2012) and Studer and Ritschard (2016). In our illustration, we only present the SA approach with an OM-metric.…”

“…The choice for either of these metrics may affect the nature of the resulting typology. So, one should be aware of the differences between these metrics (see Robette & Bry (2012), Elzinga & Studer (2015) and Studer & Ritschard (2016) for a detailed discussion of these issues). Here, we experimented with a variety of distance measures: OM with various cost settings and a sequencebased vector representation (Elzinga & Studer, 2015) with various parameter settings.…”

Research on Students' Performance in Higher Education through Sequence Analysis

“…While comparisons exist between few different metrics, using empirical data, the only study comparing a large number of metrics, using a reasoned set of artificial sequences, was made by Robette and Bry (2012). They did not try to find the best metric but "rather to unravel the specific patterns to which each alternative is actually more sensitive".…”