Multidomain Functions: Sequenzo and TraMineR Mapping

This page maps Sequenzo multidomain functions to their closest R workflows for users migrating from R. Core multidomain sequence construction is in TraMineR; linked polyadic analysis is closer to TraMineRextras.

What This Section Covers

This section follows the four multidomain strategies discussed by Ritschard et al. (2023):

• IDCD: independence from domain costs and distances
• CAT: cost additive trick
• DAT: distance additive trick
• CombT: combined typology approach

These strategies differ in what is combined: states, costs, distances, or typologies.

Quick Strategy Guide

Strategy	What is combined?	Main idea	Best used when
IDCD	States first, then costs/distances at MD level	Treat combined states as regular sequence states	You want to avoid additive assumptions and compute distances directly at the multidomain level
CAT	Domain costs	Build MD costs by additively combining domain-level costs	You want multidomain costs to reflect additive domain-level differences
DAT	Domain distances	Add separate distance matrices	You want a simple linear combination of domain-specific dissimilarities
CombT	Domain typologies	Cross-classify separately derived domain clusters	You want the joint typology to remain directly interpretable as combinations of domain-specific clusters

Mapping by Multidomain Strategy

Strategy	Sequenzo function(s)	R counterpart(s)	Notes
CAT	compute_cat_distance_matrix	TraMineR::seqMD(..., what="cost"/"diss"); older scripts may use TraMineR::seqdistmc()	Direct conceptual match for CAT-style multidomain costs and CAT-based distances.
DAT	compute_dat_distance_matrix	No direct one-function equivalent	In R this is usually done manually: compute one distance matrix per domain with TraMineR::seqdist(), then add or linearly combine the matrices outside seqdist().
IDCD	create_idcd_sequence_from_csvs + MD-level distance step	Closest workflow: build combined-state sequences with TraMineR::seqMD(..., what="MDseq"), then apply TraMineR::seqdist() directly to the resulting MD sequences with costs set at the MD level	Implemented in Sequenzo for constructing combined-state MD sequences from multiple CSV files. Distances are then computed explicitly in the next step (for example, get_distance_matrix).
CombT	get_interactive_combined_typology, merge_sparse_combt_types	No direct one-function equivalent	Usually done as a workflow: cluster each domain, cross-classify domain cluster labels, then optionally merge sparse combined types.

Supporting Functions Around the Four Strategies

Purpose	Sequenzo function(s)	R counterpart(s)	Notes
Check state association between domains (especially relevant for CAT)	get_association_between_domains	custom state cross-tabs; trajectory-level domain association in TraMineR core uses TraMineR::dissdomassoc()	A useful diagnostic before choosing a strategy. This diagnoses state co-occurrence, not trajectory association between domain distance matrices.
Linked polyads (separate framework)	linked_polyadic_sequence_analysis	Closest R package family: TraMineRextras::seqpolyads()	Related multidomain context, but not one of the four strategies in Ritschard et al. (2023).

Beginner Migration Tips

• First choose your strategy (IDCD, CAT, DAT, or CombT), then choose function(s).
• If your R script used TraMineR::seqMD(..., what="diss") with costs derived from domain-level costs, you are usually closest to CAT in Sequenzo.
• For IDCD, think in two steps: build MD sequences first, then compute MD distances.
• Use get_association_between_domains() to diagnose state co-occurrence. For DAT, trajectory association is more directly assessed by correlating domain-specific distance matrices.
• In papers/reports, always name the strategy explicitly (IDCD/CAT/DAT/CombT), because assumptions differ.

Authors

Documentation: Yuqi Liang

References

Ritschard, G., Liao, T. F., & Struffolino, E. (2023). Strategies for multidomain sequence analysis in social research. Sociological Methodology, 53(2), 288-322.