Description
Converts paired-comparison preferences into an interval scale of perceived quality or preference, with explicit model assumptions.
Listed in the treatise's borrowing matrix under psychophysics, Thurstone's law of comparative judgement models each stimulus as producing a normally distributed internal response; the proportion of times A is preferred to B maps to the distance between their means. Case V assumes equal variances and zero correlations. The output is an interval scale, so differences between items are meaningful, which ratings and rankings do not guarantee. Bradley-Terry models are a logistic analogue. In UX, paired comparisons are easier and more reliable for participants than absolute ratings, making Thurstone scaling well suited to visual design variants, voices, tones or AI response quality.
Attributes
- Discipline
- Survey & Psychometrics
- Loss type
- Unmigrated
- Stage
- Evaluation
- Data
- Quantitative
- Effort
- Days
Steps
- Select 5-15 stimuli (designs, voices, responses).
- Present all or a balanced subset of pairs, randomising order and position.
- Collect preference judgements from 20-100 participants.
- Compute choice proportions and convert to z-scores (Case V) or fit a Bradley-Terry model.
- Check model fit (circular triads, goodness-of-fit).
- Report the interval scale with confidence intervals and segment differences.
Sources
- Lost Methods in UX — A Complete Treatise (Gopinath)
- Thurstone LL (1927) A law of comparative judgment. Psychological Review 34:273-286
- Bradley RA & Terry ME (1952) Rank analysis of incomplete block designs. Biometrika
- Thurstone, A law of comparative judgment (1927), Psychological Review
- Thurstone & Chave, The Measurement of Attitude (1929)
- Torgerson, Theory and Methods of Scaling (1958)
- Feature Prioritization Methods (slide deck)
Also known as
- Paired comparison scaling
- Thurstone Case V
- Law of comparative judgement
- Equal-appearing intervals
- Thurstone Scaling
Categories
- Analytics & Behaviour
- Survey & Psychometrics