Functional measurement (information integration)

Description

Replicated factorial rating experiments whose interaction patterns reveal whether people combine attributes additively, multiplicatively or otherwise, then yield interval-scale part-worths.

Information integration theory treats a rating as a linear function of an underlying overall value. Additive, multiplicative, distributive and other combination rules each predict a specific pattern of interactions in a factorial design, so ANOVA or regression on replicated ratings can falsify a model rather than merely fit one. If the right model holds, marginal means are interval-scale estimates of part-worths. Graphs of rating curves are read first: parallel lines indicate addition, diverging fans multiplication or complements, convergence substitutes. A developmental variant compares judgement with action responses on the same factorial task and finds dissociations, for instance children whose produced speeds follow physics while their judgements do not. The method is demanding but unusually diagnostic about decision rules, which ordinary conjoint analysis assumes rather than tests.

Attributes

Loss type
Siloed
Effort
Weeks

Steps

  1. Choose two or three attributes with at least three levels each (two levels only allow additivity to be rejected).
  2. Build the complete factorial of profiles.
  3. Have each respondent rate every profile at least twice on a graphic or line scale.
  4. Plot rating curves for one attribute across levels of another and inspect for parallelism, fanning or convergence.
  5. Run repeated-measures ANOVA or per-replication regressions, testing each effect against its replication-by-effect error.
  6. Read the pattern: no interactions means additive; all interactions significant and positive means multiplicative; negative signs indicate substitutes.
  7. Estimate part-worths from marginal means, predict ratings under the chosen model and test residuals.
  8. Optionally collect action responses on the same task and compare their integration pattern with the judgements.

Sources

  • Analyzing Decision Making: Metric Conjoint Analysis (Louviere, 1988)
  • Wiley-Blackwell Handbook of Childhood Cognitive Development (Goswami, 2011)
  • Anderson, N. H. (1981). Foundations of Information Integration Theory. Academic Press

Also known as

  • Functional measurement diagnosis of decision rules
  • Functional measurement
  • Information integration theory analysis
  • Cognitive algebra testing
  • Conjoint model diagnosis
  • Functional measurement of children's intuitive physics with judgement versus action responses

Categories

  • Analytics & Behaviour
  • Behavioral Science & Economics