Академический Документы
Профессиональный Документы
Культура Документы
Title
Year
Journal
Topic
2010
Journal of the
Academy
Marketing
Science
2004
Journal of the
Academy
Marketing
Science
Understanding salespersons'
preferences for various contest
designs
2002
Journal of
Marketing
Research
2004
Journal of
Marketing
Research
2004
Journal of
Marketing
Research
A comment on the
model developed by Bradlow,
Hu, and Ho
Rao V.R.
Comments on Conjoint
Analysis with Partial
Profiles
2004
Journal of
Marketing
Research
A comment on the
model developed by Bradlow,
Hu, and Ho
Rubin D.
2004
Journal of
Marketing
Research
A comment on the
model developed by Bradlow,
Hu, and Ho
Bradlow E.T., Hu Y.
and Ho T.
Bradlow E.T., Hu Y.
and Ho T.
Haaijer R.,
Kamakura W. and
Wedel M.
Modeling Behavioral
Regularities of
Consumer Learning in
Conjoint Analysis
Incentive-Aligned Conjoint
Analysis
2004
2005
Journal of
Marketing
Research
Journal of
Marketing
Research
Journal of
2000 Marketing
Research
Journal of
2004 Marketing
Research
Journal of
2002 Marketing
Research
Journal of
Marketing
Journal of
Marketing
A cross-validity comparison of
rating-based and choicebased conjoint analysis
models
International
Journal of
Research in
Marketing
International
Journal of
Research in
Marketing
Choice of Supplier in
Wathne K. H., Biong Embedded Markets:
H. and Heide J.B. Relationship and Marketing
Program Effects
Hennig-Thurau T.,
Henning V., Sattler
H., Eggers F., and
Houston M. B.
Moore W.L.
A simple mechanism to
incentive-align conjoint
experiments
2004
2001
2004
2009
International
Journal of
Research in
Marketing
International
Journal of
Research in
Marketing
International
Journal of
Research in
Marketing
International
Journal of
Research in
Marketing
International
Journal of
Research in
Marketing
Baumgartner B.,
Steiner W.J.
Capturing consumer
Otter T., Tuchler R., heterogeneity in metric
and Frqhwirthconjoint analysis
Schnatter S.
using Bayesian mixture
models
Estimating aggregate
consumer preferences from
online product reviews
2004
2010
2008
2009
International
Journal of
Research in
Marketing
International
Journal of
Research in
Marketing
Sichtmann C.,
Stingel S.
Methodological issues in
conjoint analysis: a case
study
Jaeger S. R.,
Hedderley D. and
MacFie H. J. H.
Creusen M.,
Veryzer R. and
Schoormans J.
2007
2005
Marketing
science
Marketing
science
Marketing
science
Evgeniou T.,
Boussios C. and
Zacharia G.
Offering Online
De Bruyn A.,
Recommendations with
Liechty J., Huizingh Minimum Customer Input
E. and Lilien G.
Through Conjoint-Based
Decision Aids
Probabilistic Polyhedral
Methods for
Toubia O., Hauser J.
Adaptive Choice-Based
and Garcia R.
Conjoint Analysis:
Theory and Application
2005
2008
2008
2007
2003
Marketing
science
Marketing
science
Marketing
science
Marketing
science
Marketing
science
Preference
model
Stimulus
presentation
Part-worth
function, 6
attributes, 4-5
levels,
Part-worth
function 5
attributes, 2-3
levels,
Part-worth, 2
attributes at 2
levels,
3attributes at
3 levels, 2 at 4
levels, and 1 at
6 levels
Vector model, 6
attributes, 2
levels
Verbal
description
Verbal
description. The
learning basedmodel was
based on a
experiment
composed of
two phases:
learning
(prior) and
rating.
The authors ask for solutions to attribute density in conjoint research such as: to understand whether and how respondents
the conjoint procedure and the need for cross-disciplinary work
The authors ask for: other ways to conceptualize the problem, managerial aspects of the BHH procedure , the role of price in
or two of the previous profiles need to be complete (not partial). Issues regrding BHHs assumption of the independence of c
The author proposes the use of posterior predictive checks in evaluation of the models
They present a clarification of the original model, propose an integration of several new imputation rules add new measurem
tests. The authors also discuss general modeling challenges when researchers want to mathematically define and integrate
suggesting several critical success factors for modeling behavioral regularities in marketing. The authors encourage collabor
marketing domain itself but also across different fields (e.g., economics, operations, psychology, sociology, statistics) as a w
Fractional factorial design. S1: 3
groups of 12 choice
sets. Each choice set had 3 profiles
(Chinese meals) and
a none of the above option. The
restaurant served the meal they
chose. S2: 27 conjoint tasks, 30
unique snack combos for the holdout
task
S1: part-worth
model, 8
attributes, 2-4
levels S2: 4
attributes, 2-5
levels
Vector model, 6
attributes, 2-6
levels: brand
(6), speed (4),
technological
type (6),
digitizing
option (3),
facsimile (2),
and price (4)
Path worth, 8
attributes, 2-4
levels
Path worth, 6
product
attributes at 3
levels each
Physical
products
Verbal
description
Before respondents
answered the stated-choice questions,
they reviewed
detailed descriptions of the levels of
each feature and could
access the descriptions at any time by
clicking the features
logo. 4 sets with 8 features
Verbal
description
Verbal
description
Vector model, 4
factors each
with 2 levels
16 cards
Verbal
description
Part-worth, 5
attributes, 2-5
levels
Verbal and
pictorial
description
Verbal
description
Part-worth, 7
attributes, 3
levels
Part-worth,
7attributes,
each with 3
levels
Verbal
description
part-worth, 2
attributes, 5 and
3 levels
Verbal
description
part-worth, 2
attributes, 3 and
5 levels
Verbal
description
part-worth, 23
attributes, 2-4
levels
part-worth, CBC:
3 attributes, 3,5
levels, The HITCBC reduces the
number of levels
at two: the best
and worst level,
the authors
started the
empirical study
with 6
attributes, 3-6
levels
Verbal
description
The authors developed optimal designs for the no-choice multinomial logit model, the extended no-choice multinomial logit
optimality criterion and the modified Fedorov algorithm and compare these optimal designs with a reference design, which i
prediction accuracy. They conclude that taking into account the no-choice option when designing a no-choice experiment on
the model, used for estimation, matches the data-generating model
Linear, six
attributes of two Full profile, 133 respondents
levels each,
Verbal
description
The study shows that one criterion, Akaikes Information Criterion (AIC) with a per-parameter penalty factor of 3 (AIC3), is cle
data configurations, having the highest success rate and producing very low parameter bias. See complete article for more d
Taking into account the existance of this segmentation methods: automatic interaction detection and its multivariate variant
analysis; discriminant analysis; multidimensional scaling; conjoint analysis and componential segmentation, the authors rea
that could have been useful in practice, whereas the innovative alternative of mixture regression modelling generats segme
Part-worth, 3
attributes, 3,2,3
levels
Part-worth, 4
attributes, 2-4
levels
verbal
description
Psyhical
prototype
stimuli,
photographic
images and
verbal
description
Part-worth, 2
attributes, 3
levels
Vector, 2
attribute, 2
levels, preferred
level of each
Full profile, 422 respondents
visual design
principle (high
or low)
Part-worth, 5
attributes, 2
levels
Psyhical
products
Realistic
pictures,
pictorial model
verbal and
visual
The authors propose a new class of hierarchical dynamic Bayesian models applied to simulated conjoint data and explore th
heterogeneity across a number of possible types of dynamic effects demonstrating the derived benefits versus static models
demonstrate that using a counterbalanced design is important from an estimation perspective when parameter dynamics ar
The authors reach the conclusion that the likelihood principle is implicit to the Bayesian approach to statistics where the pos
Bayesian analysis conditions on the data to draw inferences about unobservable parameters in the analysis. In a conjoint an
I know about the part-worths?" Their view is that the answer to this question is more managerially relevant than the corresp
multiple datasets. See complete article for more details
The biases and inefficiencies are real and in the direction predicted. The authors provide stylized models and more general e
phenomena. Furthermore, empirically, they find no evidence that metric utilitybalanced questions reduce response error. Co
appears to be a more important goal than utility balance. See complete article for more details.
They reach the conclusion that their approach significantly outperforms both the method of Toubia et al. (2004) and standar
weaker when data from an orthogonal design are used. (this limitation indicates that it may be important to combine the pro
a simple method for handling heterogeneity lead to promising results with performance often similar to that of HB and estim
MNL: 1
attribute, 4
levels, CBC: 20
binary attributes
and the price
Part-worth, 5
attributes, 2,3
levels
Part-worth, 5
features at 4
levels each
4 partially
Full profile, 616 graduate and balanced blocks using an orthogonal
undergraduate students
fractional factorial
design
Verbal
description
Psyhical
products
Pictorial and
verbal
description
The method uses centrality con-cepts and ellipsoid shape approximations. The authors tested the method using a series of M
particularly suited to contexts where re-searchers are limited to asking relatively few questions compared to the number of p
found that the relative accuracy of the method is due, at least in part, to the design of the questions. Their simulations sugg
enhance existing estimation methods. See complete article for more details
Measurement
scale
Estimation
dependent
method
var.
Choice
Multinomial
logit
Rank order, 1
(the most
preferred)
Scheffe tests
to 16 (the
least
preferred)
Observations
Pretest feedback
Selfexplicated
and rating
scores. The
model
estimates a
set of scaling
constants for
each
respondent
They developt a
finite mixture
The model has
regression
important influnce
model for full
on predictive validity
profile
of CA
conjoint
Rating scale,
09 Likert
scale, choice
Hierarchical
Bayesian
approach to
account for
heterogeneity
4 as holdouts for
validation
hether and how respondents deal with missing information, to reduce density before the implementation of
rocedure , the role of price in solving the problem and a data collection procedure for partial profiles. One
tion of the independence of counts when multiple attributes are missing
tion rules add new measurement metrics for pattern matching, and draw a roadmap for further real-world
atically define and integrate behavioral regularities into traditional quantitative domains. They conclude by
e authors encourage collaborations not only between behavioral researchers and modelers within the
, sociology, statistics) as a way to undertake challenging and important research in marketing in the future
Rating 17
agree
disagree
scale, choice
Insample
hit rate and log- Out-of-sample
marginal
predictions
probability
Choice
choice
choice
They develop a
multinomial
probit (MNP)
model
HB and AC
Finite mixture,
HB models
application
ed opportunities for
ectively. The authors compared complementary methods for understanding customer-needs combinations:
Conjoint analyses, Truck clinics, Listening in. See complete article for more details.
Rating scale 1
to 16
choice
Rating scale,
010 scale
and second
study choice
Choice
pretests
Hierarchical
Bayes routine
hierarchical
Bayesian
multinomial
logit model
Additional holdout
Hierarchical
Bayesian
multinomial
logit model
choice
Hierarchical
Bayes mixture
of normals
model
estimate both
8 additional
the RCM and the
evaluations of
20-point
LCM by the
the 23 full-factorial
rating scales
Markov Chain
design were
Monte
generated as holdout
Carlo methods
profiles
ion (NBR) model is supported by an additional ACA study using the concerning attributes. This
when combining both methods to reach a more reliable estimation of the preferences existing in a market
bles the estimation of parameters, which allow inferences on the relative effect of product attributes and
products. See complete article for more details.
choice
multinomial
logit
4 additional holdout
choice sets. A
validity test shows
that this procedure
can compete with
state-of-the-art CBC
methods.
d no-choice multinomial logit model, and the nested no-choice multinomial logit model using the Dh a reference design, which is constructed while ignoring the no-choice option, in terms of estimation and
g a no-choice experiment only has a marginal effect on the estimation and prediction accuracy as long as
11-point scale
ordinary least
squares (OLS)
enalty factor of 3 (AIC3), is clearly the best criterion to use across a wide variety of model specifications and
ee complete article for more details
n and its multivariate variant; canonical analysis; factor analysis; cluster analysis; regression
egmentation, the authors reach the conclusion that traditional K-means clustering fails to produce segments
n modelling generats segments that have clear marketing strategy potential
Rank order
Choice
linear
regression
Multinomial
logit
In terms of validity,
both methods do not
show satisfactory
results for measuring
WTP.
Predicted choice
probability
Rating
Seven-point
scale ranging
from little
preference
to a lot of
preference
full ranking
Regression
model
ANOVA
Follow-up sample
conjoint data and explore the performance of these new dynamic models, incorporating individual-level
benefits versus static models. The authors also introduce the idea of an unbiased dynamic estimate, and
when parameter dynamics are present. See complete article for more details
ch to statistics where the posterior distribution is derived from the prior distribution and the likelihood.
the analysis. In a conjoint analysis, it provides an answer to the question "Given the data at hand, what do
ally relevant than the corresponding frequentist question concerning performance of an estimator across
d models and more general explanations with which to understand and isolate the cause of these
ns reduce response error. Contrary to common wisdom, orthogonality (efficiency) in metric questions
bia et al. (2004) and standard logistic regression; is less sensitive to noise, high response error; is relatively
important to combine the proposed method with a method similar in spirit for designing questionnaires) it's
milar to that of HB and estimates the interaction coefficients significantly better than all other methods
choice
MNL and
correlated probit
100-point
preference
scale
Regression
model
Choice
They used as a
comparation 4
methods: HB,
AC, ACi and Ace.
See complete
article for more
details.
follow-up question
Holdout exercise
Holouts validation
choice questions
he method using a series of Monte Carlo simula-tions. The findings confirm that the polyhedral algorithm is
compared to the number of parameters. By isolating the impact of the question design component, they
tions. Their simulations suggest that hybrid polyhedral ques-tion-selection methods could be used to
Author(s)
Title
Year
Journal
Topic
2010
Journal of the
Academy
Marketing
Science
2004
Journal of the
Academy
Marketing
Science
Understanding salespersons'
preferences for various contest
designs
2002
Journal of
Marketing
Research
2004
Journal of
Marketing
Research
2004
Journal of
Marketing
Research
A comment on the
model developed by Bradlow,
Hu, and Ho
Rao V.R.
Comments on Conjoint
Analysis with Partial
Profiles
2004
Journal of
Marketing
Research
A comment on the
model developed by Bradlow,
Hu, and Ho
Rubin D.
2004
Journal of
Marketing
Research
A comment on the
model developed by Bradlow,
Hu, and Ho
Bradlow E.T., Hu Y.
and Ho T.
Modeling Behavioral
Regularities of
Consumer Learning in
Conjoint Analysis
2004
Journal of
Marketing
Research
Bradlow E.T., Hu Y.
and Ho T.
Haaijer R.,
Kamakura W. and
Wedel M.
Incentive-Aligned Conjoint
Analysis
2005
Journal of
Marketing
Research
Journal of
2000 Marketing
Research
Journal of
2004 Marketing
Research
Journal of
2002 Marketing
Research
Journal of
Marketing
Journal of
Marketing
A cross-validity comparison of
rating-based and choicebased conjoint analysis
models
International
Journal of
Research in
Marketing
International
Journal of
Research in
Marketing
Choice of Supplier in
Wathne K. H., Biong Embedded Markets:
H. and Heide J.B. Relationship and Marketing
Program Effects
Hennig-Thurau T.,
Henning V., Sattler
H., Eggers F., and
Houston M. B.
Moore W.L.
A simple mechanism to
incentive-align conjoint
experiments
2004
2001
2004
2009
International
Journal of
Research in
Marketing
International
Journal of
Research in
Marketing
International
Journal of
Research in
Marketing
International
Journal of
Research in
Marketing
International
Journal of
Research in
Marketing
Baumgartner B.,
Steiner W.J.
Capturing consumer
Otter T., Tuchler R., heterogeneity in metric
and Frqhwirthconjoint analysis
Schnatter S.
using Bayesian mixture
models
Estimating aggregate
consumer preferences from
online product reviews
2004
2010
2008
2009
International
Journal of
Research in
Marketing
International
Journal of
Research in
Marketing
Sichtmann C.,
Stingel S.
Methodological issues in
conjoint analysis: a case
study
Jaeger S. R.,
Hedderley D. and
MacFie H. J. H.
Creusen M.,
Veryzer R. and
Schoormans J.
2007
2005
Marketing
science
Marketing
science
Marketing
science
Evgeniou T.,
Boussios C. and
Zacharia G.
Offering Online
De Bruyn A.,
Recommendations with
Liechty J., Huizingh Minimum Customer Input
E. and Lilien G.
Through Conjoint-Based
Decision Aids
Probabilistic Polyhedral
Methods for
Toubia O., Hauser J.
Adaptive Choice-Based
and Garcia R.
Conjoint Analysis:
Theory and Application
2005
2008
2008
2007
2003
Marketing
science
Marketing
science
Marketing
science
Marketing
science
Marketing
science
Preference
model
Part-worth
function, 6
attributes, 4-5
levels,
Part-worth
function 5
attributes, 2-3
levels,
Part-worth, 2
attributes at 2
levels,
3attributes at
3 levels, 2 at 4
levels, and 1 at
6 levels
They present a
clarification of
the original
model, propose
an integration of
several new
imputation rules
add new
measurement
The
authors
metrics
for ask
The
authors
ask
Vector
model,
6
for
solutions
to
pattern
for:
other
attributes,ways
2
attribute
density
matching,
and
to
conceptualize
levels
in
conjoint
draw
a roadmap
the problem,
research
such
for
further
realmanagerial
as:
to
world
tests.
The
aspects
of the
understand
authors
also
BHH procedure
,
whether
and
discuss
the role general
of
price
how
modeling
in solving the
respondents
challenges
when
problem and
a
deal
researchers
data with
collection
missing
want
to
procedure
for
information,
to
mathematically
partial profiles.
reduce
density
define
and
One
or
two
of
The
author
before
the
integrate
the
previous
proposes
the
implementation
behavioral
profiles
need to
usethe
of conjoint
posterior
of
regularities
be
completeinto
predictive
procedure
and
traditional
(not
partial).
checks
in for
the
need
quantitative
Issues
regrding
evaluation
of
crossdomains.
They
BHHs
the
models
disciplinary
conclude
bywork
assumption
of
suggesting
the
several
critical of
independence
success
factors
counts when
for
modeling
multiple
behavioral
attributes are
regularities
in
missing
marketing. The
authors
encourage
collaborations
not only
between
behavioral
Stimulus
presentation
Verbal
description
Verbal
description. The
learning basedmodel was
based on a
experiment
composed of
two phases:
learning
(prior) and
rating.
S1: part-worth
model, 8
attributes, 2-4
levels S2: 4
attributes, 2-5
levels
Vector model, 6
attributes, 2-6
levels: brand
(6), speed (4),
technological
type (6),
digitizing
option (3),
facsimile (2),
and price (4)
Path worth, 8
attributes, 2-4
levels
Path worth, 6
product
attributes at 3
levels each
Physical
products
Verbal
description
Before respondents
answered the stated-choice questions,
they reviewed
detailed descriptions of the levels of
each feature and could
access the descriptions at any time by
clicking the features
logo. 4 sets with 8 features
Verbal
description
Verbal
description
Part-worth, 7
attributes, 3
levels
Part-worth,
7attributes,
each with 3
levels
16 cards
Verbal
description
Verbal and
pictorial
description
Verbal
description
Verbal
description
part-worth, 2
attributes, 5 and
3 levels
Verbal
description
part-worth, 2
attributes, 3 and
5 levels
Verbal
description
part-worth, 23
attributes, 2-4
levels
The authors
developed
optimal designs
for the nochoice
multinomial
part-worth, CBC:
logit model, the
3 attributes, 3,5
extended nolevels, The HITchoice
CBC reduces the
multinomial
number of levels
logit model, and
at two: the best
the nested noand worst level,
choice
the authors
multinomial
started the
logit model
empirical study
using the Dwith 6
optimality
attributes, 3-6
criterion and the
levels
modified
Fedorov
algorithm and
compare these
optimal designs
with a reference
design, which is
constructed
while ignoring
the no-choice
option, in terms
of estimation
and prediction
accuracy. They
conclude that
taking into
account the nochoice option
when designing
a no-choice
Verbal
description
Taking into
account the
existance of this
TheLinear,
study shows
six
segmentation
that
one
attributes
methods: of two Full profile, 133 respondents
criterion,
levels each,
automatic
Akaikes
interaction
Information
detection and its
Criterion (AIC)
multivariate
with a pervariant;
parameter
canonical
penalty factor of
analysis; factor
3 (AIC3), is
analysis; cluster
clearly the best
analysis;
criterion to use
regression
across a wide
analysis;
variety of model
discriminant
specifications
analysis;
and data
multidimensiona
configurations,
l scaling;
having the
conjoint analysis
highest success
and
rate and
componential
producing very
segmentation,
low parameter
the authors
bias. See
reach the
complete article
conclusion that
for more details
traditional Kmeans
clustering fails
to produce
Part-worth, 3
Full profile, 179 online
segments that
attributes, 3,2,3
interviews
could have been
levels
useful in
practice,
whereas the
innovative
alternative of
mixture
regression
modelling
generats
segments that
Part-worth,
4
have
clear
attributes, 2-4
Full profile, 120 subjects
marketing
levels
strategy
potential
Verbal
description
verbal
description
Psyhical
prototype
stimuli,
photographic
images and
verbal
description
Part-worth, 2
attributes, 3
levels
Vector, 2
The
authors 2
attribute,
propose
a new
levels, preferred
class
ofof each
level
Full profile, 422 respondents
hierarchical
visual design
dynamic
principle (high
Bayesian
or low)
models applied
The
authors
to simulated
reach
thedata
conjoint
conclusion
and explorethat
the
the
likelihoodof
performance
Part-worth,
principle
is 5
these
new
attributes,
Full profile, 305 respondents
implicit
to
the2
dynamic
levels
Bayesian
models,
approach
to
incorporating
statistics
where
individual-level
the
posterior
heterogeneity
distribution
is
across a number
derived
from
of possible the
prior
typesdistribution
of
and
the effects
dynamic
likelihood.
demonstrating
Bayesian
the derived
analysis
benefits versus
conditions
on
static models.
the
to and
draw
The data
biases
authors
also
inferences
about
inefficiencies
introduce the are
unobservable
real
in the
idea and
of an
parameters
in
direction
unbiased
the
analysis.
In
predicted.
dynamic The
a
conjoint
authors
provide
estimate, and
analysis,
it
stylized
models
demonstrate
provides
an
and
that more
using a
answer
to the
general
counterbalanced
question
explanations
design is "Given
the
data
atfrom
with
which
to
important
hand,
what
do
understand
andI
an estimation
know
about
isolate
the the
perspective
part-worths?"
cause
of these
when parameter
Their
vieware
is
phenomena.
dynamics
that
the
answer
Furthermore,
present. See
to
this question
empirically,
they
complete
article
is
find
no evidence
formore
more
details
managerially
that metric
relevant than
utilitybalanced
the
questions
corresponding
reduce response
frequentist
error.
Contrary
question
to common
concerning
wisdom,
performance of
orthogonality
an
estimatorin
(efficiency)
across
metric multiple
questions
datasets.toSee
appears
be a
complete
article
more important
Psyhical
products
Realistic
pictures,
pictorial model
verbal and
visual
less sensitive to
noise, high
response error;
is relatively
weaker when
data from an
orthogonal
design are used.
(this limitation
indicates that it
may be
important to
combine the
proposed
method with a
method similar
in spirit for
designing
questionnaires)
it's a simple
method for
handling
heterogeneity
lead to
MNL: 1
promising
attribute,
results with 4
levels, CBC: 20
performance
binary
attributes
often similar
to
and
the
that of HBprice
and
estimates the
interaction
coefficients
significantly
better than all
other methods
Part-worth, 5
attributes,
The method 2,3
levels
uses centrality
con-cepts and
ellipsoid shape
approximations.
The authors
tested the
method using a
series of Monte
Carlo simulations. The
findings confirm
that the
Part-worth, 5
polyhedral
featuresisat 4
algorithm
levels each
particularly
suited to
contexts where
re-searchers are
limited to asking
relatively few
questions
compared to the
number of
parameters. By
isolating the
impact of the
question design
component,
they found that
the relative
accuracy of the
method is due,
at least in part,
4 partially
Full profile, 616 graduate and balanced blocks using an orthogonal
undergraduate students
fractional factorial
design
Verbal
description
Psyhical
products
Pictorial and
verbal
description
Measurement
scale
Estimation
dependent
method
var.
Choice
Multinomial
logit
Rank order, 1
(the most
preferred)
Scheffe tests
to 16 (the
least
preferred)
Observations
Pretest feedback
Selfexplicated
and rating
scores. The
model
estimates a
set of scaling
constants for
each
respondent
They developt a
finite mixture
The model has
regression
important influnce
model for full
on predictive validity
profile
of CA
conjoint
Rating scale,
09 Likert
scale, choice
Hierarchical
Bayesian
approach to
account for
heterogeneity
4 as holdouts for
validation
Rating 17
agree
disagree
scale, choice
Insample
hit rate and log- Out-of-sample
marginal
predictions
probability
Choice
choice
choice
They develop a
multinomial
probit (MNP)
model
HB and AC
Finite mixture,
HB models
Rating scale 1
to 16
choice
Rating scale,
010 scale
and second
study choice
Choice
pretests
Hierarchical
Bayes routine
hierarchical
Bayesian
multinomial
logit model
Additional holdout
Hierarchical
Bayesian
multinomial
logit model
choice
Hierarchical
Bayes mixture
of normals
model
estimate both
8 additional
the RCM and the
evaluations of
20-point
LCM by the
the 23 full-factorial
rating scales
Markov Chain
design were
Monte
generated as holdout
Carlo methods
profiles
choice
multinomial
logit
4 additional holdout
choice sets. A
validity test shows
that this procedure
can compete with
state-of-the-art CBC
methods.
11-point scale
Rank order
Choice
ordinary least
squares (OLS)
linear
regression
In terms of validity,
both methods do not
show satisfactory
results for measuring
WTP.
Predicted choice
probability
Multinomial
logit
Rating
Seven-point
scale ranging
from little
preference
to a lot of
preference
full ranking
Regression
model
ANOVA
Follow-up sample
choice
MNL and
correlated probit
100-point
preference
scale
Regression
model
Choice
They used as a
comparation 4
methods: HB,
AC, ACi and Ace.
See complete
article for more
details.
follow-up question
Holdout exercise
Holouts validation
choice questions
2007
International journal
of market research
Eye-tracking information
processing in choice-based conjoint
analysis
2010
International journal
of market research
2002
International journal
of market research
2005
International journal
of market research
2003
International journal
of market research
2003
International journal
of market research
2002
International journal
of market research
2004
International journal
of market research
An empirical comparison of
methods to measure willingness to
pay by examining the hypothetical
bias
2005
International journal
of market research
2006
International journal
of market research
2007
International journal
of market research
2009
International journal
of market research
2010
International journal
of market research
2009
International journal
of market research
2010
International journal
of market research
2010
International journal
of market research
2001
International journal
of market research
An investigation of country-of-origin
effect using correspondence
analysis: a cross-national context
2004
International journal
of market research
2009
International journal
of market research
2009
International journal
of market research
2002
International journal
of market research
2010
International journal
of market research
2009
International journal
of market research
2009
International journal
of market research
2007
International journal
of market research