Travel Reference
In-Depth Information
(2007) succeeded in optimizing the parameter settings of a Support Vector Machine with a
Genetic Algorithm. Their SVM-GA results turned out to be superior to the demand forecasts of
an ARIMA model and an ordinary neural network with standard backpropagation learning.
Assessment and recommendations
Three classes of analytical methods and application areas are chosen for further treatment: Scale
Development, Structural Equation Modelling and Classifi cation Techniques. These selected
fi elds and method classes have been gaining top awareness among the research community, and
they exhibit a problem-solving potential for tourism marketing intelligence not yet fully
exploited. When drafting recommendations for tourism marketing proper attention is paid to
the development in core marketing.
Scale development
Marketing research in any sector of industry, like all empirical sciences, depends on observation
and measurement. In the social sciences the aim of attaining standardized measuring instruments
is not fundamentally different from the situation in the natural sciences. However, Bond and Fox
(2001) argue that quantitative researchers in the human sciences merely focus on data analysis
while neglecting the development of criteria for measurement quality.
Measurement issues
More than 30 years ago, Jacoby (1978) criticized the market researchers' blind acceptance of
measures while ignoring quality criteria such as reliability or validity. One year later, Churchill
(1979) published a seminal paper on developing measures for behavioural constructs and
proposed a procedure for gaining reliable and valid measures. In doing so, he discusses a series of
steps starting from the specifi cation of the construct's domain to the development of norms. A
vital point Churchill (1979) raises is the suggestion to use a new sample of data for confi rming
the dimensional structures that emerged from preceding Exploratory Factor Analysis. The claim
of having confi rmed relevant dimensions cannot be made if exploratory and confi rmatory factor
analyses are applied to the same data set. Interestingly, up to now it seems that few tourism
researchers have become aware of this requirement.
Churchill (1979) outlines the superiority of multi-item measures compared to single-item
measures. He argues that single items only imperfectly capture a concept. Rossiter (2002), on
the contrary, holds that single items which have content validity are capable of properly
representing the construct of interest. He argues that in most cases multi-items are misused as
these items are inappropriately pre-tested. Rossiter (2002) specifi es his C-OAR-SE procedure
and contends that content validity is the sole reasonable and necessary quality criterion for scale
measures. More specifi cally, he states that '[. . .] construct validity and predictive validity are
inappropriate for scale evaluation, and [. . .] reliability should be regarded only as a precision-
of-score estimate for a particular application' (2000: 308). However, he admits that until now,
no empirical proof has yet been found for his assumption. Rossiter's focus on content validity
is in contrast with Churchill's argument that face or content validity is a fi rst step but '[. . .]
not the whole story' (2000: 69). According to Bagozzi (1984) and Rigdon et al . (2011) theory-
building research requires involving both the conceptual and the empirical domain as well
as correspondence rules as the latter offer valuable information on the measure's reliability
and validity.
Search WWH ::




Custom Search