We run multicollinearity diagnostics using a linear
regression model because, as Menard (2001) [36] notes,
the functional form of the model is irrelevant for the
purposes of diagnosing collinearity. All the individual
variable VIF values were below 4, indicating that
multicollinearity does not affect our results. Due to the
panel structure of our data, we used the Hausman test to
select between a fixed-effects and a random-effects
model. The test rejects the null hypothesis, indicating
that the fixed -effects estimator was consistent [37]
.
Therefore, we used panel regression with fixed effect
modeling. We enter only single interaction of CSR and
distance into regression model each time because there
may is highly linear relation among the interactions:
We run multicollinearity diagnostics using a linear
regression model because, as Menard (2001) [36] notes,
the functional form of the model is irrelevant for the
purposes of diagnosing collinearity. All the individual
variable VIF values were below 4, indicating that
multicollinearity does not affect our results. Due to the
panel structure of our data, we used the Hausman test to
select between a fixed-effects and a random-effects
model. The test rejects the null hypothesis, indicating
that the fixed -effects estimator was consistent [37]
.
Therefore, we used panel regression with fixed effect
modeling. We enter only single interaction of CSR and
distance into regression model each time because there
may is highly linear relation among the interactions:
การแปล กรุณารอสักครู่..