Civil Engineering Reference
In-Depth Information
Tabl e 7
Summary of goodness of fit statistics for SEM of KOC, KS and IM
Goodness of fit statistics
Data-fit criterion
Results
Fit or not
2
Model overall fit
(1)
χ
The smaller
407.41
Not fit
the better(
p
>
0
.
05)
(
p
<
0
.
00)
(2) GFI
>
0
.
90
0.94
Fit
(3) AGFI
>
0
.
90
0.92
Fit
(4) RMR
<
0
.
05
0.03
Fit
(5) RMSEA
(a)
0.05 (Perfect)
(b) 0
.
05
∼
0
.
08 (Good)
<
0.069
Fit
(c) 0
.
09
∼
0
.
10 (Ok)
(Good)
(d)
>
0
.
10 (Bad)
(6) ECVI
0
∼
1, and the smaller
0.49
Fit
the better
Model comparison fit
(1) NFI
>
0
.
90
0.98
Fit
>
.
(2) NNFI
0
90
0.98
Fit
(3) CFI
>
0
.
90
0.97
Fit
>
.
(4) RFI
0
90
0.98
Fit
(5) IFI
>
0
.
90
0.99
Fit
Model parsimony fit
(1) PNFI
>
0
.
50
0.80
Fit
(2) PGFI
>
0
.
50
0.66
Fit
(3) CAIC
The smaller the better
653.55
(a)
<
Independent CAIC
<
826
.
79
Fit
(b)
<
Saturated CAIC
<
24931
(4) CN
>
200
250.44
Fit
Fig. 3
Assumed mediator
variable model
Mediator
KS
Predictor
KOC
Criterion
IM
more easily the model is unfit. The samples of the research are up to 967 so that
chi-square(
2
) result is acceptable. To sum up, it is concluded that SEM of KOC,
KS, and IM is proved to have a good model data-fit.
χ
5. KS, though with a weak mediation effect, is proved a mediator variable.
To judge whether KS is a mediator variable between KOC and IM, the mediator
variable model (
Bagozzi and Yi 1988
) is used in this study (see Fig.
3
). First, Route
A coefficient should reach the significant level. Second, Route B coefficient should
Search WWH ::
Custom Search