To perform our analysis we used a Partial Least Squares and Structural Equation Modeling tool (PLS-GRAPH Version 3.00 build 279). SEM allows researchers to simultaneously examine the structural component (path model) and measurement component (factor model) in the one model.

We performed our validation by constructing two PLS models, one for each application. The internal consistency (reliability) statistics for all constructs in both models were above 0.96. This exceeded the 0.7 rule-of-thumb [26] C. Fornell and D. Larcker, Evaluating structural equation models with unobservable variables and measurement error, Journal of Marketing Research 18 (1981), pp. 39–50. Full Text via CrossRef[26] and confirmed the scales’ reliability. We tested convergent validity by examining whether all items loaded highly on their respective construct in PLS. A common rule-of-thumb is a loading greater than 0.7 [63]. In the email and word processor models, all items loaded on their constructs from 0.87 to 0.96, indicating convergent validity. To test discriminant validity, we tested the item-to-total correlations for each system. As shown in Table 3 and Table 4, the items loaded cleanly on each construct, indicating good discriminant validity.