Unfolding Model
Unfolding Models for Preference Data · Also known as: Ideal Point Model, Preferential Choice Scaling, Coombs Unfolding, Katlanma Modeli
The Unfolding Model is a geometric approach to preference analysis that represents both individuals and choice objects (stimuli) as points in a shared low-dimensional space. Originating with Clyde Coombs's foundational 1950 work on preferential choice and rigorously systematized by Borg and Groenen (2005), the model assumes each person prefers the stimulus closest to their personal ideal point, thereby 'unfolding' rank-order preference data into a joint spatial map.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use the Unfolding Model when you have rank-order or rating preference data and wish to visualize individuals and choice objects in a common geometric space. The method suits consumer preference studies, political ideology scaling, and psychometric attitude research. Key assumptions include Euclidean distance as the preference metric and a unimodal (single-peaked) preference function per respondent. It is not appropriate for dominance data or when preferences are non-unimodal. Alternatives include Multidimensional Scaling for similarity data or Bradley-Terry models for paired comparisons.
Strengths & limitations
- Jointly maps both respondents and stimuli in a single interpretable space, revealing preference segments at a glance.
- Handles rank-order data without requiring interval-level measurement assumptions.
- Flexible dimensionality allows trade-offs between parsimony and fit quality.
- Geometrically transparent: preference predictions follow directly from distances, making results easy to communicate to non-technical audiences.
- Assumes a single-peaked (unimodal) preference function, which may not hold when respondents have complex or lexicographic preferences.
- Solution is not unique; rotations, reflections, and translations of the coordinate system yield equivalent fits, requiring external anchoring for interpretation.
- Sensitive to the number of dimensions chosen: too few distort structure, too many overfit.
- Large numbers of missing rank judgments can destabilize the estimation, as every respondent–stimulus distance contributes to the Stress function.
Frequently asked
How does the Unfolding Model differ from classical Multidimensional Scaling?
Classical MDS starts from dissimilarity (or similarity) data between stimuli only and produces a map of stimuli. The Unfolding Model additionally incorporates individual respondent ideal points derived from preference rankings, yielding a joint map of both people and stimuli. The input data type — preferences rather than pairwise dissimilarities — is therefore fundamentally different.
What Stress value indicates an acceptable unfolding solution?
Following conventions in multidimensional scaling literature (Borg & Groenen, 2005), Stress values below 0.05 are considered excellent, 0.05–0.10 good, 0.10–0.15 fair, and above 0.20 poor. However, Stress depends on data size and dimensionality, so these thresholds should be used as guidelines rather than absolute cutoffs, and the researcher should also inspect the resulting configuration for substantive interpretability.
Can the Unfolding Model handle rating scales instead of full rank orders?
Yes. When preference intensities are recorded on a rating scale (e.g., 1–7 Likert items), metric unfolding can be applied, treating the ratings as approximate distances. Nonmetric unfolding uses only the ordinal information in the ratings. Both variants are described by Borg and Groenen (2005); the choice depends on whether the analyst is willing to assume the rating scale values carry interval-level meaning.
Sources
- Borg, I., & Groenen, P. J. F. (2005). Modern Multidimensional Scaling: Theory and Applications (2nd ed.). Springer. ISBN: 978-0-387-25150-9
How to cite this page
ScholarGate. (2026, June 2). Unfolding Models for Preference Data. ScholarGate. https://scholargate.app/en/statistics/unfolding-model
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Bradley-Terry ModelDecision-making↔ compare
- Correspondence AnalysisStatistics↔ compare