As described in the development of the measurement model for the original Model
5, there is a clear item content overlap between EM4 = I have a great deal of
control over my job and EM5 = I am given responsibility at work. In the latest
model (SEM 5b:1) the standardised loading estimate for EM4 is 0.787 and for EM5
it is 0.709.
From the standardised residual covariance matrix, the total unmeasured variance
associated with EM4 is 7.5 and for EM5 this value is 10. With a lower loading and
a greater level of associated unmeasured variance, EM5 is removed.
Construct Item Standardised factor loading estimates
ML EM EPA JP DSB
Motivational
Leadership
ML1 (^) .904
ML2 .953
ML3 .878
Employee
Empowerment
EM (^1) .749
EM3 (^) .728
EM4 (^) .784
EM5 (a)^ .708
Employee Positive
Attitudes
JS1 .740
ME1 .877
ME3 .849
ME6 .80 3
AOC2 .771
Job Performance
JP1 (^) .861
JP2 (^) .865
JP4 .458
Discretionary
Service Behaviour
DSB1 (^) .929
DSB2 .698
Average
variance
extracted (AVE)
0.768 0.607 0.700 0.618 0.725
Construct
reliability (CR) 0.908^ 0.861^ 0.921^ 0.820^ 0.838^
Model fit
statistics
χ^2 = 125.093; d.f. = 109; sig = 0.139
RMSEA = 0.026 (0.046; 0.000; pclose = 0.982)
CFI = 0.992
SRMR = 0.0386
CN (0.05) = 228
(a) (^) Item removed in the structural model
Table 7-29 Estimates and model fit for CFA 5b:8
Following the removal of EM5, the model is re-estimated as SEM 5b:2. The
parameter estimates and model fit statistics are all satisfactory and are illustrated
in Figure 7 - 22. Prior to discussing these, however, because the factor structure of
the measurement model has been modified (the indicator EM5 has been