RESULTSMeasurement ModelThe first step in our analysis was to test the fitof the measurement model. Model fit isassessed in terms of three indices: comparativefit index (CFI), goodness-of-fit index (GFI) androot mean square of approximation (RMSEA).A model is considered to be satisfactory if CFI> 0.95, GFI > 0.90 and RMSEA < 0.06 (Hu andBentler 1999; Bearden et al. 1993). The firstmeasurement model tested did not fit the datawell [x" (38)=71.27, CFI=0.94, GFI=0.89,RMSEA=0.09). A closer look at the LISRELoutput revealed that several of the measurementitems for Green Lifestyle were correlated witheach other. The measurement model wastherefore refined to allow these measures tocorrelate. The resulting model exhibitedsatisfactory fit ({^ (35)=32.80, CFI=1.00,GFI=0.94, RMSEA=O.OO).In addition to model fit, we examined theconvergent and discriminant validity of themeasurement items for each latent variable.Table 2 summarizes the results of this analysis.Convergent validity refers to the extent towhich multiple items measuring the sameconstruct are in agreement (Nunnally 1978),and was assessed three ways. First, thestandardized loading factors, which indicate thelevel of agreement between measurement itemsand a latent variable, are all significant(p<0.001) for the one multi-measured latentvariable. Green Lifestyle. Second, the intemalconsistency for the measurement items wascalculated using the composite reliability score
developed by Werts and colleagues (1973).
Composite reliability should be interpreted like
a Cronbach's alpha coefficient and should
exceed 0.70 (Fomell and Larcker 1981).
Finally, the average variance extracted (AVE)
is the ratio of the construct variance to the total
variance among the indicators, and should be
greater than 0.50 (Fomell and Larcker 1981).
The composite reliability and AVE values in
Table 2 exceed recommended levels and thus
the latent variable of Green Lifestyle
demonstrates good convergent validity.
41
Florenthal andArling
Discriminant validity refers to the extent to
x'viiich a construct is different from other
constructs. Constructs demonstrate discriminant
validity if the AVE is higher than the squared
correlation between the constructs (Fomell and
Larcker 1981). The square root of the AVE of
the Green Lifestyle construct (0.75) is higher
than the correlations between the other
constructs, demonstrating discriminant validity.
การแปล กรุณารอสักครู่..
