262 / 372
工作要求與工作控制的交互作用對倦怠與學習努力之影響：公平知覺之干擾角色
262
3.4 Burnout
Burnout was measured with four items adapted from a scale developed by Singh (2000)
to measure burnout among sales employees. The alpha coefficient is 0.87.
3.5 Fairness Perception (FP)
The FR was measured with five items adapted from a measure used by Janssen (2001).
The alpha coefficient is 0.91.
3.6 Control Variables
We controlled the salespeopleʼs gender, age, education, and tenure due to their potential
relevance to the independent and dependent variables, based on past research (e.g., De Rijk
et al., 1998; Wall et al., 1996; Xie, 1996).
4. Results
To assess the convergent and discriminant validity of all measures, a measurement
model of all multi-item measures was subjected to confirmatory factor analysis. The overall
fit statistics for our seven-factor model indicate an acceptable fit to the data: χ
2
(188, N =
148) = 323.05,
p
< 0.01; CFI = 0.91; IFI = 0.91; RMSEA = 0.069. To assess the discriminant
validity of the factors in the measurement model, two kinds of analyses were conducted.
First, according to suggested procedures of Bagozzi et al. (1991), we conducted a series of
confirmatory factor analyses to test whether, for each pair of factors in the measurement
model, a two-factor model have a significantly better fit than a one-factor model. The results
of the chi-square difference test showed that the two-factor model fit the data significantly
better than the one-factor model. Secondly, following the procedures suggested by Fornell
and Larcker (1981), we found that the average variance extracted for two constructs exceed
the square of the correlation between the constructs. The findings reveal that all constructs
showed sufficient discriminant validity.
We employed procedures described by Podsakoff et al. (2003) to exclude the influence
of common method bias because our data were collected from a single source. We did this
even though interaction term effects, which are at the center of this research, are not affected
by such a bias (Evans, 1985; Schmitt, 1994). In addition, we conducted the Harman’s one-
factor test to address common method variance (Podsakoff et al., 2003). The findings
showed that a single method-driven factor does not adequately represent our data and that
our results are unaffected by common method bias.