Results Validity Stage (Modified Model)
For testing the modified model, it was specified in LISREL with data taken from sample 4. Several scholars recommend that if model modifications were made during the scale development and refinement process, the final model should be cross-validated by using a new sample (e.g., MacKenzie et al., 2011; Netemeyer et al., 2003; Worthington & Whittaker, 2006). While this rarely is done in scale research, the study at hand collected new data (sample 5(a-e); n = 824) from Austrian online consumers in order to confirm scale reliability and convergent validity. This data was gathered with a series of street-intercept interviews, throughout using a paper-and-pencil questionnaire. It was expected that the construct of eWOM trust is likely to percist in alternative population.
Model validity was supported, as the solution was proper, all hypothesized relationships were statistically significant, none of the items showed negative error variances, and the completely standardized parameter estimates were all greater than one. The purified model showed a satisfactory fit. While the chi-square statistic was significant (x2 = 627.59 (822.45), df = 204, p
< .001) the absolute, incremental and parsimonious fit measures suggested that the proposed second-order model, with its five first-order constructs measured by 22 observable indicators adequately represented the structure within the sample data (GFI = .90 (.89), AGFI = .88 (.85), RMSEA = .057 (.064), RMR = .09 (.11), SRMR = .06 (.07); CFI = .95 (.92), NNFI = .95 (.91), NFI = .93 (.91); normed chi-square: 3.1 (4.0), PIF = .73 (.71)).