Page - 44 - in Critical Issues in Science, Technology and Society Studies - Conference Proceedings of the 17th STS Conference Graz 2018

Scale quality The analysis of tau-equivalent reliability shows that all relevant (and expressive) values for Cronbach’s alpha are situated between 0,709 and 0,864, which was considered to be a sufficient internal consistency for those items, where an examination of internal consistency is methodologically sensible. Some scales though, such as negative experience, while measuring one construct, by design have no need for internal consistency. For instance, negative experience is a formative construct, which means, it is formed by several individual negative experiences which are substantially different but still form one construct. Since both, factor analysis and measurements of internal consistency, aim at correlated items, these tests are not applicable for this type of constructs (cf. Eberl 2004, p. 7). App usage on the other hand is a manifest construct directly measured by items that simply retrieve the usage frequency of different app types. Hence, it is self- explanatory why this scale does not feature any internal consistency. The scale for Reputation qua factor analysis turned out to contain two separate factors, one was neglected. Due to this fact, the scale was reduced to three items which is a possible explanation for the low alpha value and shows a need for scale improvement before the representative survey. Figure 8: Analysis results of the internal consistency SEM After the inspection of scale quality, the examined constructs were operationalised due to the results of this very inspection. The so formed indices were inserted into a structural equation model (SEM), which was then reduced within multiple iterations. The SEM is needed to model 44