|
Latent
Class Analysis (LCA) is a statistical
method used to categorize individuals into distinctive groups
or classes. Unlike traditional methods, LCA takes into account
both observed and unobserved variables. Consequently, it can
yield improvements over traditional approaches to cluster,
factor, regression/segmentation, and neural network applications,
and related graphical displays. It's also advantageous because
it does not rely on the traditional modeling assumptions,
which are often violated in practice, such as assumptions
of linearity, normal distribution, and homogeneity. In addition,
indicator variables do not have to be standardized, and can
be used in conjunction with other measures using different
scales. Thus, it is less subject to biases associated with
data that do not conform to model assumptions.
Applications:
- Exploratory Analysis
- Market Segmentation
- Development of Attitudinal or Behavioural
indexes
- Estimating the Measurement Error of
Attitudinal or Behavioural indexes
- Measuring Customer Brand Loyalty
- Predicting Brand Switching Behaviour
- Measuring Consumer Preferences
- Predicting Consumer Buying Behaviour
Advantage over
Traditional Cluster / Factor Analysis:
Traditional cluster / factor analysis requires
variables to be linear, and have a similar variation and
scale. In addition, these traditional data reduction methods
do not allow for the simultaneous inclusion of covariates.
Accommodating covariates requires a two stage analysis,
such as using discriminate analysis, which may lead to misclassification,
and other errors.
When using Likert-type measures, such as
strongly disagree, disagree, agree, and strongly agree,
as a means of clustering individuals into similar attitudinal
or behavioural groups, rarely do these measures conform
to the assumptions of linearity. In most cases they simply
are ordinal, in that there is an order of magnitude as one
moves along the scale, from strongly disagree to disagree,
and from disagree to agree, and so on. But linearity also
means that the spacing within the scale is the same. In
other words, the space between strongly disagree and disagree
is the same as the space between disagree and agree, and
between agree and strongly agree. Since this assumption
is rarely met, the results may be in accurate. The accuracy
of LCA does not depend on meeting these assumptions. Attitudinal
measures can be nominal, or ordinal, and the scale and variance
can differ. There is also no assumption of linearity, or
equal spacing within the measurement scale.
Finally, in traditional clustering procedures,
convention and ad-hoc guess-work are used to determine the
number of clusters. Since LC is based on a statistical model,
statistics are available to help determine the number of
clusters.
Advantages
over Multiple Regression:
Traditional regression assumes
homogeneity across an entire population, which does not allow
for the existence of different segments. LC or mixture regression
involves estimating a regression model under the assumption
that the regression coefficients differ across unobserved
(latent) segments, yielding improved predictions. LCA regression
can also be used in conjunction with repeated measures, such
as in conjoint analysis
Advantages
over Logistic and Multinomial Regression:
Similar to multiple regression,
categorical regression also benefits when the model parameters
are allowed to vary by latent class. Doing so may reveal,
for example, that a car's purchase price is related to individuals'
desire to purchase a new car versus a pre-owned car. In other
words, the assumption is that price is important to everyone.
This may not necessarily be the case. For some individuals,
price may be extremely important, while for others it may
have no bearing on their decision to purchase a new or pre-owned
car. Latent class regression allows price, or any other indicator,
to vary by latent class or behavioural group, and thus individuals
with similar attitudes or behaviours can be clustered or grouped
together. |
Latent Class Analysis
does not assume linearity, normal distribution
of data, or homogeneity of variances.
TYPES
OF LATENT CLASS ANALYSES:
- LATENT CLASS CLUSTER ANALYSIS
- LATENT CLASS FACTOR ANALYSIS
- LATENT CLASS MULTIPLE REGRESSION
- LATENT CLASS LOGISTIC / MULTINOMIAL
REGRESSION
- LATENT CLASS LOG-LINEAR MODELS
(latent class structural equation models)
- LATENT CLASS DISCRETE CHOICE
MODELS
What
you will learn:
- Conducting latent class cluster,
factor, regression analysis.
- Interpreting model results
- Imposing model restrictions
- Creating latent classes and importing
these classes into traditional analyses, such as regression
models, log-linear, or discrete choice models
- You will learn how to fit latent
class models using LEM, LATENT GOLD, and Mplus
2-Class
Regression Model:
Advantage Over Other
Courses or Workshops:
Unlike other
courses teaching latent class analysis, we do not use only
one software package, since this workshop is not a means of
marketing software. This course will be taught using all the
leading software packages: LEM, Latent Gold, and Mplus. Each
software package has its advantages and disadvantages, and
limitations. |
