This is, y = dependant variable (transportation services), x = independent
variable (commitment), z = moderation variables (internal orientation and
market orientation) and xz = added variables (commitment * internal
orientation and commitment * market orientation).
Research on a moderating effect is to test the effect of moderating
variables on the relationship between independent and dependant variables.
A precondition for MRA is that independent variables must have causal
links on dependant variables. After testing that, this research can analyze
whether moderating variables have a moderating effect on dependent
variables. The process of the analysis is as follows. First step compares
Model II with Model III after estimates of three models. R2 of Model III
must be higher than R2 of Model II and b3 must be not 0 (zero) and this
explains that internal orientation and market orientation as moderating
variables have a moderating effect on the relationship between
commitment and transportation services. In this regard, the moderating
variable is a pure moderating variable if b2 is 0 but the variable is regarded
as a quasi-moderating variable if b2 is not 0 because there is a moderating
effect. If there is no moderating effect, it appears that R2 of Model III is
lower than R2 of Model II and b3 is 0.
Non-response bias is tested by the method recommended by Armstrong
and Overton. The sample firms are divided into four groups in arrival
order. The annual turnover of the firms is compared to the first group and
the last group. If there is no gap in the two groups, there is no problem in
non-response bias in the sample firms. The result of analysis of variance is
1.416 in F value and 0.237 in p value. This means that there is no problem
in non-response bias in the sample firms.