latent constructs which are identified by factor analysis (Klem,
2000). In this study, from a measurement CFA, a structural
path model was developed and tested with the maximum
likelihood procedure in SPSS/AMOS. Also, an attempt was
made to accommodate the influence of moderating variables
(such as age, experience, and culture) on behavioral intention
(BI) and use behavior (UB). In this model, the v
2 is 418.585
with 328 degrees of freedom. The v
2
/df ratio is equal to 1.27.
The other fit indexes are, respectively, RMSEA = 0.030, RMR
= 0.031, CFI = 0.985, NFI = 0.935, RFI = 0.926 and other
indicators that indicate a good fit. The final structural path
model is shown in Figure 7.4.
The results of Figure 7.4 are summarized in Table 7.6.
The table depicts the summary of path coefficients with their
regression weights, SE, CR, and significant (p) values.
It is evident from Table 7.6 that the standardized
regression coefficient (SRC) from FC to UB is 0.034 (with
CR = 0.447 and p = 0.655) is a non-significant link.
Therefore, the link FC ! UB was removed from the final
structural model and the model was rerun as recommended
by Venkatesh, Thong, & Xu (2012).
It is important to note that the structural model (without
considering the moderation impact) showed that R square for
BI is 0.58 that explains 58% of the model, whereas R square
of the UB is 0.235 that explains the 23.5% of the model. In
other words, five independent variables explained 58% of
the variance in the instructors’ behavioral intentions to adopt
the LMS in their face-to-face teaching.
Hence, the results from Figure 7.4 and path coefficient
table of the final structural path model are summarized as:
1. Path PE to BI: The standardized regression coefficient
(SRC) from PE to BI is 0.246 CR = 4.71, and p<0.01,
which indicates a positive and significant relationship.
2. Path EE to BI: The SRC from EE to BI equals 0.198
with CR = 3.340 and p<0.01, which indicates a positive
and significant relationship between EE and BI.
3. Path SI to BI: The SRC from SI to BI equals 0.144 with
CR = 2.567 and p<0.01, which indicates a positive and
significant relationship between SI and BI.
4. Path FC to BI: The SRC from FC to BI equals 0.270
with CR = 4.338 and p<0.01, which indicates a positive
and significant link between FC and BI.
5. Path M to BI: The SRC from M to Bi is 0.214 with CR
= 3.929 and p<0.01, which indicates a positive and significant link.
6. Path BI to UB: The SRC from BI to UB is 0.459 with
CR = 5.68 and p<0.01, which indicates a positive and a
significant relationship between BI and UB.
7.6 Tests for Moderation
In this study, age, experience, awareness, and cultural
dimensions are treated as a moderator between independent
variables (PE, EE, SI, FC, HM, and H) and dependent
variable (BI). The procedure for applying moderation tests is
adapted from Im, Il Hong & Seongtae Kang, (2011) and
Gaskin (2012). In this study, the responses were collected in
more than one category. To calculate the moderating effect
in the SPSS/AMOS program, the first median of all moderators was calculated and then based on their median, all
categories were converted into two groups. For instance, the
responses for age were collected into seven categories,
ranging from under 30 years to 60 years and above. Based
on their median, all seven categories of age were converted
into two groups, named as older age and younger age
groups. A similar approach was applied to convert all other
moderators into two groups such as low experience and high
experience. In this test, the z value indicates the significance
of the mean difference between the two groups (Byrne,
2010). The moderation testing process includes two main
tests: (a) model-level moderation test and (b) path-by-path
moderation test.
7.6.1 Model-level Moderation
In the moderation test, the model is run for two groups (low
and high) without any constraints (i.e. free model) and the
chi-square (v
2 ) is noted. Then the model is constrained by
setting the ‘equality constraint’ for the same groups and the
model is run again. The chi-square (v
2 ) is again noted. The
chi-square (v
2 ) values of both models are compared. Using
the ‘Stats Tools Package’ of Gaskin (2012), the chi-square
(v
2 ) values, along with their degrees of freedom (df), and the
p-values are calculated for overall moderation (see Stats
Wiki and Stats Tools Package (http://statwiki.
kolobkreations.com/index.php?title=Main_Page)).
The
results of the moderation test at model-level are summarized
in Table 7.7.
7.5 Model Testing Procedures
67
2000). In this study, from a measurement CFA, a structural
path model was developed and tested with the maximum
likelihood procedure in SPSS/AMOS. Also, an attempt was
made to accommodate the influence of moderating variables
(such as age, experience, and culture) on behavioral intention
(BI) and use behavior (UB). In this model, the v
2 is 418.585
with 328 degrees of freedom. The v
2
/df ratio is equal to 1.27.
The other fit indexes are, respectively, RMSEA = 0.030, RMR
= 0.031, CFI = 0.985, NFI = 0.935, RFI = 0.926 and other
indicators that indicate a good fit. The final structural path
model is shown in Figure 7.4.
The results of Figure 7.4 are summarized in Table 7.6.
The table depicts the summary of path coefficients with their
regression weights, SE, CR, and significant (p) values.
It is evident from Table 7.6 that the standardized
regression coefficient (SRC) from FC to UB is 0.034 (with
CR = 0.447 and p = 0.655) is a non-significant link.
Therefore, the link FC ! UB was removed from the final
structural model and the model was rerun as recommended
by Venkatesh, Thong, & Xu (2012).
It is important to note that the structural model (without
considering the moderation impact) showed that R square for
BI is 0.58 that explains 58% of the model, whereas R square
of the UB is 0.235 that explains the 23.5% of the model. In
other words, five independent variables explained 58% of
the variance in the instructors’ behavioral intentions to adopt
the LMS in their face-to-face teaching.
Hence, the results from Figure 7.4 and path coefficient
table of the final structural path model are summarized as:
1. Path PE to BI: The standardized regression coefficient
(SRC) from PE to BI is 0.246 CR = 4.71, and p<0.01,
which indicates a positive and significant relationship.
2. Path EE to BI: The SRC from EE to BI equals 0.198
with CR = 3.340 and p<0.01, which indicates a positive
and significant relationship between EE and BI.
3. Path SI to BI: The SRC from SI to BI equals 0.144 with
CR = 2.567 and p<0.01, which indicates a positive and
significant relationship between SI and BI.
4. Path FC to BI: The SRC from FC to BI equals 0.270
with CR = 4.338 and p<0.01, which indicates a positive
and significant link between FC and BI.
5. Path M to BI: The SRC from M to Bi is 0.214 with CR
= 3.929 and p<0.01, which indicates a positive and significant link.
6. Path BI to UB: The SRC from BI to UB is 0.459 with
CR = 5.68 and p<0.01, which indicates a positive and a
significant relationship between BI and UB.
7.6 Tests for Moderation
In this study, age, experience, awareness, and cultural
dimensions are treated as a moderator between independent
variables (PE, EE, SI, FC, HM, and H) and dependent
variable (BI). The procedure for applying moderation tests is
adapted from Im, Il Hong & Seongtae Kang, (2011) and
Gaskin (2012). In this study, the responses were collected in
more than one category. To calculate the moderating effect
in the SPSS/AMOS program, the first median of all moderators was calculated and then based on their median, all
categories were converted into two groups. For instance, the
responses for age were collected into seven categories,
ranging from under 30 years to 60 years and above. Based
on their median, all seven categories of age were converted
into two groups, named as older age and younger age
groups. A similar approach was applied to convert all other
moderators into two groups such as low experience and high
experience. In this test, the z value indicates the significance
of the mean difference between the two groups (Byrne,
2010). The moderation testing process includes two main
tests: (a) model-level moderation test and (b) path-by-path
moderation test.
7.6.1 Model-level Moderation
In the moderation test, the model is run for two groups (low
and high) without any constraints (i.e. free model) and the
chi-square (v
2 ) is noted. Then the model is constrained by
setting the ‘equality constraint’ for the same groups and the
model is run again. The chi-square (v
2 ) is again noted. The
chi-square (v
2 ) values of both models are compared. Using
the ‘Stats Tools Package’ of Gaskin (2012), the chi-square
(v
2 ) values, along with their degrees of freedom (df), and the
p-values are calculated for overall moderation (see Stats
Wiki and Stats Tools Package (http://statwiki.
kolobkreations.com/index.php?title=Main_Page)).
The
results of the moderation test at model-level are summarized
in Table 7.7.
7.5 Model Testing Procedures
67