Description
Respondents rate each attribute's levels and each attribute's importance separately, and the two are multiplied into part-worths without evaluating whole products.
Self-explicated measurement is the compositional alternative to conjoint analysis: instead of inferring part-worths from choices among whole profiles, respondents state them piece by piece. They rank or rate the levels within each attribute, then rate how important the difference between the best and worst level would be if products were otherwise equal. Level desirability multiplied by attribute importance gives prior part-worths. Because attributes are never traded off together, Orme's glossary says such models are not conjoint analysis. They serve as priors in Adaptive Conjoint Analysis, in hybrid methods combining priors with trade-off questions, and for imputing utilities of levels not shown in choice tasks. Orme warns that importance grids are answered in about five seconds, crowd the top of the scale, discriminate poorly and invite social desirability (understating price); dropping them from ACA improved predictions under hierarchical Bayes.
Attributes
- Discipline
- Survey & Psychometrics
- Loss type
- Siloed
- Goal
- Prioritise, Measure attitudes
- Stage
- Strategy, Definition
- Data
- Quantitative
- Effort
- Days
Steps
- Define attributes and levels, checking that attributes do not overlap (overlapping attributes are double-counted).
- Have respondents rank or rate the levels within each attribute.
- For each attribute ask how important the difference between its most and least preferred level is (for example on a 1-4 scale).
- Multiply level desirability by attribute importance to form prior part-worths per respondent.
- Optionally combine with paired comparisons or choice tasks (hybrid conjoint) or use the priors to impute utilities for levels not shown.
- Check predictive validity against holdout choices before using results for decisions.
Sources
- Getting Started with Conjoint Analysis (Orme, 2010)
Also known as
- Self-explicated approach
- Compositional preference measurement
- Self-explicated conjoint
- Attribute importance rating
Categories
- Analytics & Behaviour
- Survey & Psychometrics