working in a relatively homogenous service context (restaurants in commercial
hotels).
The research design follows a positivistic organisational psychology /
organisational behaviour approach wherein relevant variables and psychometric
methods of quantifying these are identified from reviews of the relevant literature.
The key variables in this research are latent (unobservable) constructs (also
referred to as latent variables, latent factors, or factors) which are operationalised
and measured using multiple indicator variables. Structural equation modelling
(SEM) is chosen as the analytical method because this technique makes it
possible to measure relationships between latent variables. Further advantages
of SEM include its capacity to (i) estimate models containing multiple dependent
and independent variables and (ii) to incorporate mediating variables (see e.g.
Raykov and Marcolides 2006: 7).
Analysis of data is undertaken using IBM SPSS and AMOS (the IBM SPSS module
for SEM analysis; both SPSS and AMOS Version 18) using structural regression
models (Kline 2005: 209). The modelling method is based on Anderson and
Gerbing’s (1988) two-step procedure (measurement model followed by structural
model) with each step following the five stages (specification, identification,
estimation, testing and modification) described by Schumaker and Lomax (2004).
The research employs a Model Generating (MG) approach in which the model is
modified and tested again using the same data (Jöreskog 1993: 295; Raykov and
Marcolides 2006: 7). Whilst structural equation models are specified and tested
in a hypothetico-deductive manner – that is, each causal or correlational
relationship is posited in the form of an a priori hypothesis which is then tested
statistically – the Model Generating or theory development approach means that
at a broader level the research operates in a somewhat exploratory mode. On
this matter, Raykov and Marcolides write:
In theory development, repeated applications of SEM are carried out,
often on the same data set, in order to explore potential relationships
between variables of interest. In contrast to the confirmatory mode of
SEM applications, theory development assumes that no prior theory
exists—or that one is available only in a rudimentary form—about a
phenomenon under investigation.
(Raykov and Marcolides 2006: 6-7)