Data Analysis
Before testing our model, we conducted a preanalysis and data validation according
to the latest standards, for six purposes: (1) to determine if the indicators were formative
or reflective and to properly model the first-order and second-order factors; (2) to
establish the factorial validity of the measures by examining convergent and discriminant
validity; (3) to establish that multicollinearity was not a problem with any of the
measures; (4) to check for common methods biases; (5) to establish reliability; and
(6) to test the manipulations that we added to provide variation in the model. These
results are available on request of the authors. Table 2 summarizes our measurement
model statistics and average variance extracted (AVE) analysis. We used partial least
squares (PLS) regression via SmartPLS version 2.0 [89] for model analysis because
PLS is especially adept at validating mixed models of formative and reflective indicators
and because component-based structural equation modeling (SEM) techniques
are more appropriate for theory development than are covariance-based techniques
[7, 20, 88]. To test our model, we generated a bootstrap with 500 resamples. Table 3
summarizes the results of the tested paths for the hypotheses, rival explanations, replications,
and covariates. Figure 2 depicts the significant paths and R2s.
Data AnalysisBefore testing our model, we conducted a preanalysis and data validation accordingto the latest standards, for six purposes: (1) to determine if the indicators were formativeor reflective and to properly model the first-order and second-order factors; (2) toestablish the factorial validity of the measures by examining convergent and discriminantvalidity; (3) to establish that multicollinearity was not a problem with any of themeasures; (4) to check for common methods biases; (5) to establish reliability; and(6) to test the manipulations that we added to provide variation in the model. Theseresults are available on request of the authors. Table 2 summarizes our measurementmodel statistics and average variance extracted (AVE) analysis. We used partial leastsquares (PLS) regression via SmartPLS version 2.0 [89] for model analysis becausePLS is especially adept at validating mixed models of formative and reflective indicatorsand because component-based structural equation modeling (SEM) techniquesare more appropriate for theory development than are covariance-based techniques[7, 20, 88]. To test our model, we generated a bootstrap with 500 resamples. Table 3summarizes the results of the tested paths for the hypotheses, rival explanations, replications,and covariates. Figure 2 depicts the significant paths and R2s.
การแปล กรุณารอสักครู่..