Travel Reference
In-Depth Information
tandem with other advanced analytical methods and examples will be mentioned in the
next section.
There are typical examples of how to combine analytical methods in tourism marketing
research. Regression techniques and Exploratory Factor Analysis are the methods to enter into
such combinations most frequently. EFA lets the analyst condense a number of observables
into 'factors' (or principal components) and benefi t from their orthogonality in subsequent
regression runs. Such two-step setups are still in use. As a typical example see Molina and Esteban
(2006) who apply PCA and logistic regression. However, with the advent of SEMs there is no
more need for proceeding in a stepwise manner when analyzing cause-effect relationships
among a set of hypothetical constructs. Single-step methods are not just superior because of
statistical reasons. They enforce a unifi ed view of theory building acknowledging that the
measurement and structural submodels are constituents of an integral theory and must not be
tested and adjusted separately in an incremental approach to model fi tting.
Classifi cation procedures are common in method combinations. For example, Bargeman,
Timmermans and Van der Waerden (1999) search for profi ling criteria of tourist segments. The
tourist clusters are derived out of panel data with the Sequence Alignment Method and the cor-
relates of segment membership are detected with a double-application of decision-tree analysis
(CHAID) and a loglinear model. Further processing a cluster structure of tourists is not limited
to simple profi ling with passive variables. As Jang, Morrison and O'Leary (2004) demonstrate,
the selection of target segments may be optimized in terms of receipts and seasonal stability by
means of Quadratic Programming.
Multiple uses of methods are sometimes driven by the analysts' desire to cross-validate
empirical fi ndings. This is particularly helpful where a method does not include statistical
signifi cance testing. For example, Fodness and Milner (1992) run a connected sequence of four
analytical steps:
1 aggregating theme park visitation patterns into a 'visitor interchange matrix';
2 processing these park similarity data with MDS;
3 clustering visitors by their individual visitation history into market segments; and
4 testing a number of demographic and socio-economic variables for profi ling the visitor
segments.
Dale Fodness in a 1994 article moves from MDS to PCA and on to a Partitioning Clustering
method when measuring tourist motivation. Molera and Albaladejo (2007) make joint use of
factorization of perceived benefi t items with principal components, hierarchical and partitioning
clustering, ANOVA testing of cluster differences, and multinomial logit regression for segment
profi ling.
Smart combinations of methods are not limited to the traditional toolkit. Tsaur, Tzeng and
Wang (1997) worked on assessing tourist risks. These authors relied on Fuzzy Set Theory, but
facing a situation of multi-criteria decision making elegantly linked it to the Analytical Hierarchy
Process. Doing this they avoided the artifi cial accuracy of crisp measurements while aggregating
the tourists' risk evaluation criteria. As tourism has begun uniting forces with Information
Technology new opportunities for incorporating Artifi cial Intelligence methodology have
opened. A promising fi eld is the development and refi nement of travel recommender systems.
Fuzzy Reasoning is one of the instruments for making these systems smarter (Franke 2003).
There is an ample fi eld of incorporating analytical methods in travel counselling systems. As an
example consider Wallace
et al
.'s (2004) proposal of feeding hierarchically clustered trip plans into
a Radial Base Function Neural Network for achieving rapid online response. Chen and Wang
