estimates and model fit diagnostics for the current model iteration (CFA 5:7) are
described Table 7 - 26.
Before translating CFA 5:7 into the structural model, a final check to confirm
discriminant validity is made. This reveals that the constructs do not demonstrate
discriminant validity – specifically, the lowest AVE (EM = 0.582) does not exceed
the highest squared correlation estimate of 0.602 (for EPA EM).
No further modifications are suggested by the range of diagnostic information
(factor loadings, the standardised residual covariance matrix or the modification
indices). The next stage, therefore, is to check and remove any outlier cases
(following the guidelines for using the Malahanobis distance values (D^2 ) provided
by Byrne 2010: 105-106) and then to re-estimate the measurement model to see
if discriminant validity has improved.
Construct Item Standardised factor loading estimates
ML WV EM EPA JP DSB
Motivational
Leadership
ML1 .903
ML2 .954
ML3 .878
Work Values
WV 1 .728
WV4 .761
WV5 .814
WV6 (a)^ .837
Empowerment
EM1 .769
EM3 .718
EM5 .655
Employee Positive
Attitude
JS1 .740
AOC2 .773
ME1 .87 4
ME3 .850
ME6 .803
Job Performance
JP1 .882
JP2 .846
JP4 .451
Discretionary
Service Behaviour
DSB1 .908
DSB2 .714
Average
variance
extracted (AVE)
0.768^ 0.721^ 0.582^ 0.700^ 0.617^ 0.716^
Construct
reliability (CR) 0.908^ 0.911^ 0.806^ 0.921^ 0.818^ 0.833^
Model fit
statistics
χ^2 = 186.157; d.f. = 155; sig = 0.045
RMSEA = 0.031 (0.046; 0.005; pclose = 0.954)
CFI = 0.987
SRMR = 0.0424
CN (0.05) = 211
(a) Item removed in the structural model