Nevertheless, (i) all of the indicator variables have statistically significant loadings
on the factors, (ii) the variance extracted (AVE) values for the latent factors are
all acceptable at ≥0.5, and (iii) the construct reliability (CR) values are all more
than satisfactory at ≥0.08.
Construct Item Standardised factor loading estimates
(^) ML JP DSB
Motivational
Leadership
ML1 .898^
ML2 .943^
ML3 .889^
ML4 .602^
ML5^ .606^
Job Performance
JP1 .847
JP2 .860
JP3 .556^
JP4^ .494^
Discretionary
Service Behaviour
DSB1 .798^
DSB2 .775^
DSB3 .677^
DSB4^ .693^
Average variance
extracted (AVE) 0.560^ 0.557^ 0.585^
Construct
reliability (CR) 0.860^ 0.826^ 0.849^
Model fit statistics
χ^2 = 203.726; d.f. = 62; sig = 0.000
RMSEA = 0.104 (0.120; 0.088; pclose = 0.000)
CFI = 0.9 05
SRMR = 0.0725
CN (0.05) = 85
Table 7-1 Estimates for the measurement model CFA 1:1
Finally, looking at the factor loadings for each indicator variables, we can see that
4 items are flagged for attention:
ML4 and ML5 lie around the lower bounds of the 0.6 to 0.7 range; and
JP3 and JP4 are both below 0.6.
DSB3 and DSB4 are close enough to 0.7 not to warrant attention for the time
being.
The model meets the requirements for nomological validity insofar as the latent
factors all covary at a statistically significant level (p = <0.001) and in the
hypothesised direction (in this case, all in a positive direction).
The process of model modification towards improving the model fit is now
described. Throughout the model development process described in Sections 7.2