Conjoint Analysis
1. Definition
▢ Conjoint analysis is the technique that could predict the goods chosen by
consumers, which estimates the goods' attributes and utility as they
represent methodology analyzing the consumer’s worth.
▢ Conjoint analysis is a set of techniques ideally suited to studying
customer’s choice processes and determining trade-offs (Rao, 2010).
2. Model types
▢ Conjoint analysis is classified by the part-worth function model, vector
model, and ideal point model depending on the methodology estimated
level-worth by attributes.
❍ The part-worth function approach has received wide acceptance
due, in part, to the ready interpretability of the graphically
displayed attribute part-worth functions.
source: 이훈영(2021)
❍ We can set the model that considered the main effect in the
part-worth model.
❍ Where is the number of attributes, n is the number of profiles,
is the number of levels of j-th attribute, and is the
indicated variable defining the level of each attribute.
❍ If we got the value of
, we can get the part-worth( ) of t-th
level of j-th attribute
for
for
❍ Moreover, we can the relative importance of each attribute.
max min
▢ The method of data collection is classified by full profile method and
trade-off method.
source: 이훈영(2021)
3. Steps involved in conjoint analysis
Step Alternative methods
1. Selection of a model Vector model, idea-point model, part-worth
of preference function model, mixed model
Two-factor-at-a-time (trade-off analysis),
2. Data collection method
full-profile (concept evaluation)
3. Stimulus set
Fractional factorial design, random sampling from
construction for the
multivariate distribution
full-profile method
Verbal description (multiple cue, stimulus card),
4. Stimulus presentation paragraph description, pictorial or
three-dimensional model representation
Paired comparisons, rank order, rating scales,
5. Measurement scale for
constant-sum paired comparisons, category
the dependent variable
assignment
MONANOVA, PREFMAP, LINMAP, Johnson’s
6. Estimation method nonmetric tradeoff algorithm, multiple regression,
Logit, Probit