The fourth and final step in the measurement model is the assessment of discriminant validity. Rouf and Akhtaruddin (2018) suggested that a measurement model should ensure discriminant validity. The discriminant validity is defined as “the extent to which a construct is empirically distinct from other on structs in the structural model (Hair et al., 2019, p. 9). According to Farrell (2010), “researchers cannot be certain whether results confirming hypothesized structural paths are real or whether they are a result of statistical discrepancies” (p. 324). Discriminant validity has been assessed using Fornell -Larcker criterion and cross-loadings (Hair et al., 2011).
The first criterion for the establishment of discriminant validity is the method of cross-loadings (Hair et al., 2016). Rule of thumb to assess cross-loadings approach is “indicator’s loading with its associated latent construct should be higher than its loadings with all the remaining constructs (i.e., the cross-loadings)” ( Hair et al., 2011; p. 146). If cross-loadings are higher than the indicators loading, it indicates the problem of discriminant validity in the data (Hair et al., 2016).
Table 4.8 shows that the outer loadings of all constructs are exceeding the cross-loadings in each column. Therefore, discriminant validity has been verified.
The second criterion of the establishment of discriminant validity is the Fornell-Larcker criterion (Hair et al., 2016). Fornell and Larcker (1981) proposed an approach to test discriminant validity, which has been widely used in social science research. This approach suggests that “discriminant validity is established if, for each of two constructs, the squared multiple correlations between items and constructs (i.e. the average variance extracted (AVE)), is greater than the squared correlation between constructs (i.e. the shared variance (SV))” (Franke & Sarstedt, 2019; p. 431). Table 4.9 reveals the results of discriminant validity through the criteria of Fornell-Larcker. Table 4.9 revealed that the square root of each construct’s AVE (given in the diagonal in the bold form) is greater than the highest correlation with any other construct. Therefore, discriminant validity has been well established.
In recent times, Henseler, Ringle, and Sarstedt (2015) proposed an alternative measure of discriminant validity: heterotrait-monotrait (HTMT) ratio of the correlations. According to Franke and Sarstedt (2019), the HTMT approach is more comprehensive and easy to use to assess discriminant validity, especially for a researcher who applies PLS-SEM in their research. Henseler et al. (2015) suggested the threshold value of HTMT as 0.90 if constructs are conceptually similar and 0.80 if constructs are conceptually different. If HTMT values are higher than 0.90, it shows that the discriminant problem is present in the data (Hair et al., 2019). Results show that the HTMT values of all constructs are less than 0.90 (see table 4.10). Hence, there is no discriminant problem in the data.
Overall, these results show that all the measures used in this research possess adequate validity. Therefore, further analysis of hypotheses testing could have proceeded.  oglasi

Leave a Reply