- model estimation;
- model testing; and
- model modification.
Reporting of estimates
Throughout the analysis and discussion sections, standardised parameter
estimates will be used. Unstandarised estimates are reported in Appendix III to
allow for future comparative research and following the advice of Kline (2005:
318 - 319). Appendix III also reproduces the correlation matrices for the finalised
models according to Boomsma’s (2000: 470) recommendations.
6.4.1 Building the models
Following the model generating approach, the models were developed in an
iterative manner wherein the core hypotheses are firstly examined and,
thereafter, successive constructs are added to the model and the model re-
assessed to evaluate the role of each additional construct. This iterative approach
to the model development process was adopted for a variety of reasons.
Firstly, the core hypotheses (in this case H 1 to H 4 which test the relationships
between motivational leadership, work meaning and job performance/work
motivation) could be tested in the absence of additional and potentially
confounding constructs.
Secondly, as will be reported in Chapter 7 , some of the constructs were not able
to be included in the models owing to either a lack of discriminant validity or an
absence of statistically significant covariance with other constructs. By building
the models in an iterative manner, it was possible to efficaciously identify such
constructs that do not covary with the other constructs in the model.
Thirdly, the iterative approach allows for a clearer appreciation of the role that
each construct plays in the model. For example, the transition from Model 4 to
Model 5b revealed that the effect of employees’ work values on attitudes to work
is reduced to a non-significant level when the employee empowerment construct
was added to model. In this example, had these constructs been included