6 METHODS FOR DATA COLLECTION AND ANALYSIS
This research aims to explore and evaluate the contribution of motivational
leadership to employee work motivation in hospitality services and to do this
within the broader organisational / motivational context by also measuring how a
number of non-leadership phenomena contribute to employee work motivation.
Structural equation modelling provides a method to empirically examine and
evaluate these inter-relationships.
Following a general introduction to SEM, this chapter goes on to discuss the
chosen modelling approach, the specific type of SEM analysis that will be
performed and details of how the models will specified, estimated and tested.
The data requirements for SEM analysis inform the survey design in general and
the design of the survey instrument in particular. This chapter also describes the
concomitant development of the survey instrument and refinement of the
measurement scales following which the procedure for the administration of the
survey instrument is detailed.
A more detailed description of, and justification for, the specific methods
employed in the SEM analyses is provided in Chapter Error! Reference source
not found. alongside the development of the first SEM model. The development
of the subsequent models is then reported in a less exhaustive manner with new
techniques introduced as appropriate.
6.1 Structural equation modelling
Structural equation modelling (SEM) is a multivariate statistical procedure for
providing insights into the relationships between phenomena. SEM is particularly
useful when the phenomena of interest cannot readily be measured using a single
variable – that is, when dealing with multifaceted phenomena that can be more
accurately measured (and more adequately represented) using a range of
variables rather than just one. SEM achieves this by using the confirmatory factor
analysis (CFA) technique to measure these unobservable (latent) variables based
on the ‘effect’ that the latent variable has on the observable (indicator) variables.
Latent variables can also be referred to as factors, or constructs, and indicator
variables are also referred to as items.