dicated the factors were positively and significantly correlated (p <0.05). AMOS V16.0.1 the confirmatory factor analytic model was estimated via the Maximum Likelihood method (Fig. 1 ). with the structural equation modeling literature (Chen, Curran, Bollen, Kirby, & Paxton, 2008; Fan & Sivo, 2005) a range of indices were used to asses model fit. The measurement model demonstrated broadly satisfactory levels of fit across all samples. Across the calibration validation and full samples, GFI and CFI were 0.93, 0.92, 0.95 and 0.96, 0.96, 0.97 respectively. The RMSEA was 0.05 for all samples (Table 4).
Table 4 Measurement model fit indices
? 2 df? 2/dfB-S pGFIAGFIIFITLICFIRMSEA (90% CI) C127.27 801.590.160.930.900.960.950.960.05 (0.04-0.07) V133.96801.680.080.920.880.960.950.960.05 (0.04-0.07) F162.36 802.030.020.950 .930.970.960.970.05 (0.04-0.06)=calibration sample (n=211), V=validation (n=210) and F=full sample (n=421) item loadings on their hypothesized dimensions for the calibration, validation and full samples ranged from 0.62 to 0.95, 0.50 to 0.98 and 0.55 to 0.97 respectively. The smallest t=6.7 (p <0.001) represented highly significant item loadings across all samples. Across all three samples the average dimension correlation was r=0.57 with the smallest t=3.77. This indicated a strong inter-action effect between dimensions.assess the internal consistency of the five B2B service brand identity subscales the validation sample was used to calculatecomposite reliabilities. All composite reliabilities exceeded the recommended level of 0.6; employee & client focus (0.89), corporate visual identity (0.77), brand personality (0.75) consistent communication (0.73) and human resource initiatives (0.75). scale «s construct validity was assessed next. Given the item pool had been subject to expert panel review the scale was considered face valid whilst the domain of each dimension had been extensively sampled to enhance the scale »s content validity. Three tests demonstrated convergent validity. First, all average variance extracted values ??were greater than 0.5. Second, the smallest item test statistic was greater than 1.96 (t=6.7; a=0.001). Third, all standardized factor loadings were greater than 0.5. Discriminant validity was assessed in three ways. First, the x2 for the unconstrained five factor model was significantly lower than the x2 of each constrained model. Second, all estimated confidence intervals (+ / - two standard errors) for each dimension correlations did not contain the value 1. Finally, with the exception of the brand personality and consistent communications correlation, the variance extracted estimates were greater than the square of the correlations for each pair of dimensions. However, the previous discriminant validity tests show that forcing a four factor model results in a deterioration of model fit whilst there are no theoretical reasons for assuming brand personality and marketing communication-related items should be combined. The result of this analysis was a five dimensional, 15 item B2B service brand identity scale (Table 5).
5 Final B2B service brand identity scale
Employee & client focusOur organization treats each employee as an essential part of the organization Our employees will help clients in a responsive manner Our organization makes an effo...
