leadership and motivation in hospitality

Once again, the Competencies sub-factor and its indicators (EM6, EM7 and EM8)
are removed and the measurement model is re-developed with one EM construct
(relating to Hancer and George’s Influence sub-factor).

Following steps 1 to 7 as described in Table 7 - 28 , CFA5b:8 is estimated and
produces good model fit. Importantly, the new model demonstrates discriminant
validity with the lowest AVE value (0.607 for EM) exceeding the highest squared
multiple correlation value (0.545 for EM  EPA).

The parameter estimates and model fit statistics are described in Table 7 - 29.
With a satisfactory measurement model, the structural model (SEM 5b:1) is now
specified according to Figure 7 - 21. The structural coefficients are all statistically
significant however the χ^2 p value is just short of the 0.05 threshold (χ^2 p =
0.049). An inspection of the standardised residual covariance matrix reveals no
SRCs greater than 2.58. The modification indices do, however, indicate that EM4
and EM5 share some unmeasured variance and therefore account for some of the
model mis-fit. It is theoretically plausible to remove one of these items since they
measure similar components of the Empowerment construct.

Step Diagnostic observations and actions (Model 5b)

1 Step 1 is the samML4 ML5 JP3 ME5 ME7 JS2 JS3 JS4 WV2e as for Model 5, based on low factor loadings: are removed

2

EM9 loading has dropped to 0.614 – EM9 removed
SRCs highlight DSB4/JP4 at 3.499 - DSB4 removed (based on the usual
substantive reasons for retaining JP4)
3 SRCs highlight DSB3/JP4 at 2.694 SRCs also indicate that JS5/ML3 (at 2.579) should be –^ DSB3^ removed^ flagged for observation

4 SRC value for JS5/ML3 now at 2.584; JS5 is the weaker variable (loading = 0.680 versus 0.879) - remove JS5

5

Modification indices highlight AOC2/AOC4 as the largest MI; there is item

content overlap (as described in Model 3); AOC2/4 loadings are similar; AOC4

has greater associated unmeasured variance in the SRCM than (12.8 versus

7.1 for AOC2) - AOC4 removed

6

Modification indices highlight ME1/ME3 as the largest MI; however, it is

problematic to remove one of these; there is no item content overlap and

these two indicators embody two key substantive components of the EPA

factor (satisfaction and enjoyment, which are derived from Wollack et al.’s

Intrinsic Meaning domain)

The next (theoretically plausible) high MI value AOC1/AOC2; these items have

clear item content overlap (as described in Models 3 and 4); as with Models 3

and 4, AOC1 is found to perform less well than AOC2 (higher associated

SRCs) - AOC1 removed

7

Modification indices highlight JS1/JS6 as the indicator pairing with the largest
MI that it is theoretically plausible to address by making a modification; as
with previous models, JS6 has item content overlap with ME1 – JS6 removed
8 Good fit

Table 7-28 Modification steps for CFA 5b:1 to 5b:8