013 levels
1
2
3
CHOICE-BASED CONJOINT
What makes people choose?
Build a study that reveals the trade-offs behind a decision.
Define your attributes
Start with three attributes and three levels. Add more as your study grows.
023 levels
1
2
3
033 levels
1
2
3
Record expected attribute power
A planning hypothesis only. It does not weight cards or replace measured importance.
How this prototype models choice
Balanced random full-profile cards, effects coding, and aggregate multinomial logit with an L2 penalty of 0.5. Only complete sessions are analysed. Attribute importance is its utility range divided by the sum of all ranges. Choice probabilities use exp(total utility) divided by the sum across available alternatives.
No interactions, hierarchical Bayes, segmentation, significance tests or D-efficient optimisation. Repeated choices are not independent people. Formal power depends on expected effects, design and subgroup analysis.
Sources: Sawtooth: estimating utilities with logit · Sample size issues for conjoint analysis