Leisure Time-Use Sequence Analysis

Adding up minutes collapses the day into totals and throws away order, so two people with identical totals but completely different daily rhythms look the same. Sequence analysis keeps the day as an ordered string and measures similarity by how much editing is needed to turn one day into another, so it can distinguish a dawn-run-then-evening-TV day from an all-afternoon-screen-then-late-training day even when their totals match. The price is added complexity and dependence on a cost scheme, but the gain is a holistic, timing-aware comparison that duration tallies cannot provide.

There is no universally correct answer, which is why cost-setting is the method's central debate. Equal substitution costs treat every state swap as equally dissimilar and are a defensible neutral default. Theory-driven costs let you encode that, say, swapping sport for active leisure is less of a difference than swapping sport for sleep. Data-driven schemes derive costs from observed transition rates between states. Whatever you choose, report it explicitly and check whether your typology is robust to reasonable alternatives, because the distances -- and therefore the clusters -- can shift with the costs.

They are useful summaries, not natural kinds. Clustering partitions a continuous space of pairwise distances into groups for interpretability, but individuals near cluster boundaries could plausibly belong to a neighboring type, and a different number of clusters would tell a slightly different story. Treat the typology as a vocabulary for describing recurrent patterns of leisure days and for relating those patterns to covariates, and validate it with cluster-quality diagnostics and substantive interpretability rather than presenting the types as fixed categories that exist independently of the analysis.