关于测量模型的评价.docVIP

关于测量模型的评价 Table 2 briefly describes five exogenous constructs and two endogenous constructs along with the number of indicators examined for the measurement model. The results of the initial CFA indicated that all indicator loadings were significant (p 0.001), which implied that all t-values exceeded 2.58 and thus the indicators on each latent construct were considered valid. The results of reliability values on each construct also indicated that all the constructs exceeded the recommended values: “job characteristics” = 0.87, “participative decision-making” = 0.93, “pressure/stress” = 0.93, “intrinsic motivation” = 0.90, “supervisory leadership” = 0.95, “job satisfaction” = 0.84, and “intent to stay” = 0.91 (Swanson Horridge, 2004; Diamantopoulos Siguaw, 2001). To assess if the measurement model was statistically acceptable five different fit measures, as presented in Table 3, were selected and reviewed (Hair, Anderson, Tatham, Black, 1998; Kline, 1998). They included χ2 statistic, χ2/df (normed χ2), GFI (goodness-of-fit index), CFI (comparative fit index), and RMSEA (root mean square error of approximation). The results showed that the first measurement model was not fully satisfied with p = 0.000, GFI = 0.76 and CFI = 0.88, χ2/df = 2.5, and RMSEA = 0.08. Therefore, to better fit the data, four indicators (JC5, JC6, PD1, and SL5) based on the suggestion of modification indices were removed since they were too closely interrelated. Then, the second measurement model was run to see if the fit of the model improved and it was found to be relatively acceptable: χ2/df = 2.5, CFI = 0.90, and RMSEA = 0.08. The resulting model was then tested to assess if the causal relationships would exist by specifying the direct paths between the exogenous variables and the endogenous variables. Validation of the Model 模型的确认 The obtained model was found to be fairly acceptable because three of the model fit measures (χ2/df, CFI, and RMSEA) were satisfied, but the χ2 statistic was