臺大管理論叢 NTU Management Review VOL.29 NO.1

37 NTU Management Review Vol. 29 No. 1 Apr. 2019 5. Results 5.1 The Structural Model To alleviate the structural model fit influenced by the sample size, we adopted these four indices. The goodness-of-fit index (GFI) should normally be close to 0.90, the adjusted GFI (AGFI) should exceed 0.80. The root mean square residual (RMR) and the standard root mean square residual (SRMR) should be less than 0.05, and with exceptions values less than 0.08 are acceptable (Jöreskog and Sörbom, 1993; Byrne, 1998; Hu and Bentler, 1999; Diamantopoulos and Siguaw, 2000). The total fit of the structural model is adequate: GFI = 0.93, AGFI = 0.85, RMR = 0.048 and SRMR = 0.045. 5.2 Hypotheses Testing Table 3 shows the results in SEM analysis, the cause-effect relation among latent variables which include direct and indirect influences (moderate). We adopted the maximum likelihood estimation method which is well suited to assess the model parameters via the covariance matrix (Jöreskog and Sörbom, 1993). Table 3 Results When Considering Indirect Influence Hypothesis Construct path Affected latent variable Supplier hoarding intention Direct ( t -value) Indirect ( t -value) H1 Supplier belief in future earnings → Supplier hoarding intention .43 ( 9.27 ) H2 Supplier attitude towards weakened operational shock- wave after disaster → Supplier hoarding intention -.47 ( -6.09 ) H3 Perceived relationship quality in main business partner → Supplier hoarding intention -.31 ( -6.17 ) H1a Supplier belief in future earnings × Regime intervention → Supplier hoarding intention -.21 ( -8.16 ) H2a Supplier attitude towards weakened operational shock- wave after disaster × Regime intervention → Supplier hoarding intention .24 ( 6.04 )