demonstrate statistically significant correlations with the other constructs in the
model (Social Support and Service Quality); (b) did not demonstrate discriminant
validity within the wider model (Job Satisfaction, Affective Organisational
Commitment and Work Meaning^21 ); or (c) did not exert a statistically significant
effect on any other constructs (Work Values). This process is described in Section
8.2 below.
Following the evolutionary development of the models, a smaller set of eight
hypotheses that had been satisfactorily tested remained. These hypotheses are
described in Table 8 - 1 below along with a further two hypotheses relating to
mission clarity that are not included in the organising framework.
Objectives 3 and 4 were concerned with the development (based on interrogation
of existing literature and expert judging) and refinement (by means of the pre-
test and pilot surveys) of the measurement scales for the latent variables
illustrated in Figure 8 - 1.
Research Objective 5, to test and evaluate the relationships between the latent
variables, was accomplished using data collected using a questionnaire survey of
hotel restaurant waiting staff. The data collection process and assessment of the
data are described in Sections 6.5 and 6.6 while Section 8.2 below summarises
the model development process highlighting the significant modifications to, and
important findings from, each iteration of the model. The contributions of the
research, new conceptualisations and implications for research and for practice
are then discussed in Section 8.3.
8.2 The model development process
Data were collected from waiting staff in hotel restaurants in the UK and analysed
using structural equation modelling (SEM). A model generating (MG) approach
was utilised in which models are developed through iterative modification and re-
testing with the same data (Jöreskog 1993: 295; Raykov and Marcolides 2006:
7).
21
The Work Meaning construct performed satisfactorily in the absence of the other
two attitudinal constructs (Model 2) but in later models (Models 3 to 5b), the three
attitudinal constructs were consolidated into one construct owing to a lack of
discriminant validity between them.