leadership and motivation in hospitality

somewhat speculatively) and the broader substantive consequences of the
removal of EM-C are entirely manageable since the construct of core interest in
this research is EM-I (Influence) as described in Section 5.2.

Once again, for brevity, the modifications leading to the final measurement model
specification are summarised and described in Table 7 - 25.

Step Diagnostic observations and actions (Model 5)

1

The usual indicators load low on their factors

For the usual substantive reasons, JP4 is retained, though loading only 0.491

ML4 ML5 JP3 ME5 ME7 JS2 JS3 JS4 WV2 are removed

2

EM9’s loading has dropped from 0.640 to 0.616 – EM9 removed

WV3 (at 0.653) is flagged for observation

Standardised residual covariance matrix (SRCM) highlights high levels of

unmeasured variance between DSB3/JP4 DSB4/JP4 and JS5/ML3

In addition, DSB3, DSB4 and JS5 all load at <0.700

DSB3, DSB4 and JS5 removed

3

WV3 remains only moderate at 0 .654

No issues highlighted in the SRCM

Check of modification indices reveals:

 AOC2/AOC4 has the highest MI value - AOC4 removed (higher SRCs

than AOC2)

 JS6/JS1 and JS6/AOC1 both have high MI values – JS6 removed (JS6

has item content overlap with ME1)

 WV3/WV4 has a high MI – WV3 removed (WV4 has a higher factor

loading of 0.749 against WV3’s 0.654)

4

All indicators load satisfactorily (except JP4) and all SRCs are <2.58

Modification indices highlight AOC1/AOC2 as the largest MI where there is

item content overlap (as described in Model 3); AOC1/2 loadings are similar;

AOC1 has greater associated unmeasured variance in the SRCM (14.5 versus

9.1 for AOC2) – AOC1 removed

5

All indicators load satisfactorily (except JP4) and all SRCs are <2.58

Modification indices highlight WV4/WV7 with highest theoretically plausible MI

value; no item content overlap; factor loadings are very similar; WV7 has

highest associated unmeasured variance in the SRCM (WV4 = 15.5; WV7 =

19.6)

WV7 removed

6

All indicators load satisfactorily (except JP4) and all SRCs are <2.58
Modification indices highlight EM4/EM5 with highest theoretically plausible MI
value; there is a clear item content overlap (EM4 = I have a great deal of
control over my job; EM5 = I am given responsibility at work); EM4 has lower
factor loading (0.705 versus 0.777) and higher levels of associated
unmeasured variance in the SRCM (11.1 versus 5.2)
EM4 removed
7 Good fit (almost)

Table 7-25 Modification steps for CFA 5:1 to 5:7

Following these 6 modification steps, the constructs all demonstrate content
validity (AVEs >0.5 and CR values >0.7) and model fit is reasonable (the only
issue is the χ^2 p value at 0.045, just below the 0.05 level). The parameter