Travel Reference
In-Depth Information
their seminal marketing research paper introducing heterogeneity into SEM, Jedidi, Jagpal and
DeSarbo (1997) argue that though a step-by-step approach is useful if segments can be identifi ed
a priori, it is not satisfactory for detecting unobserved heterogeneity. The frequent habit of
considering data collected for the purpose of testing a model to be drawn from one single popu-
lation may produce misleading results. Additionally, standard goodness-of-fi t measures cannot
detect unobserved heterogeneity. They propose a fi nite mixture structural equation model where
path estimates and unobserved heterogeneity are treated simultaneously. Accordingly, probabilis-
tic clusters are formed and cluster-specifi c estimates for the structural and the measurement
models are obtained in parallel. Finite mixture modelling should be used 'when substantive
theory supports the structural equation model, a priori segmentation is infeasible, and theory
suggests that the data are heterogeneous and belong to a fi nite number of unobserved groups'
(Jedidi et al . 1997: 39). This approach has not yet been widely applied, even though it has been
identifi ed to be promising for market segmentation (Steenkamp and Baumgartner 2000).
New methodology for treating heterogeneity non-parametrically is waiting for applications,
too. A non-parametric approach is often more appropriate due to the rarely met distributional
assumption of parametric analysis. A combined Latent Class and Vector Quantization approach
to perceptions-based market segmentation is discussed in Mazanec and Strasser (2007).
Within SEM, nonlinear relationships and tackling the intricate problem of causality with
novel instruments of inferred causation theory are promising areas for future applications
(Mazanec 2007a). Reaching out into the bigger latent variable family, Latent Class Analysis
and Latent Growth Models hold promise as well. Latent Growth Models have only recently
found their way into the tourism fi eld, however, not yet into tourism marketing research.
Consequently, tourism research has just scratched the surface of latent variable analysis. Readers
interested in an overview of latent variable modelling that exceeds ordinary SEMs are referred
to Muthén (2002).
Classifi cation techniques
Marketing strives to classify market segments to be capable of catering to the needs of homog-
enous groups more precisely. The ultimate goal of segmentation and accordingly customized
tourist products, services and marketing mix actions is gaining competitive advantage.
Classifi cation approaches
Two basic approaches of classifi cation have been used (Bailey, 1994). The fi rst a priori (Smith
1989: 46ff.) or common-sense segmentation (Dolnicar 2004a) is theory driven, meaning
segments are known in advance. The second a posteriori (Mazanec and Strasser 2000) or data-
driven approach (Dolnicar 2004a) is explorative; no prior information about segments is available.
Among the most popular methods for a posteriori classifi cation is cluster analysis. Cluster analysis
groups individuals based on the similarity of response patterns regarding selected variables by
minimizing within-group variance and maximizing between-group variance (Wedel and
Kamakura 2000).
Cluster analysis methods
The family of methods used for performing a posteriori segmentation comprises hierarchical
clustering (agglomerative or divisive), iterative partitioning clustering, density search, factor-
analytic procedures, clumping (permitting overlapping clusters), graph theoretic algorithms and
Search WWH ::




Custom Search