Figure 7-17 Structural specification for Model
Before estimating the measurement model using the now complete data set, it is
worthwhile briefly describing the double-headed arrow connecting Motivational
Leadership (ML) and Work Values (WV) in Figure 7 - 17. Model 4 is the first model
in this research to include more than one exogenous construct – that is, both ML
and WV are independent latent variables that have no causes contained within the
model (unlike the dependent, endogenous latent variables, that are each
hypothesised to be influenced by another one or more latent variables in the
model).
In SEM it is usual to assume that exogenous variables are correlated, unless there
is a specific theory-based rationale arguing against this (Garson 2011b). Kline
(2005: 67-68) describes this correlation as representing the unanalysed
association between the two constructs, going on to say that the correlation is
unanalysed in the sense that no predictions are offered to explain why these
constructs covary.
The procedure for measurement model development used in for Models 1 to 3 is
followed once more. For brevity, the modifications leading to the final
measurement model specification are summarised in Table 7 - 22.
Following these 6 modification steps, the measurement model now demonstrates
satisfactory fit and all factor loadings and factor covariances are statistically
significant (see Table 7 - 23 ). The constructs all demonstrate convergent (all AVEs
0.5 and all CRs > 0.07) and discriminant validity (the highest squared
correlation estimate is ML EPA at 0.325 and the lowest AVE is 0.617).