Log in / Register
Home arrow Engineering arrow Measuring Electronic Word-of-Mouth Effectiveness: Developing and Applying the eWOM Trust Scale

Results Reliability Stage

As mentioned earlier, in the reliability stage, initial data were gathered by setting an online questionnaire to 850 students of two major Austrian universities (University of Vienna and Vienna University of Economics and Business). Given the fact that the sample was sufficiently large, it was possible to split the data into two sub-samples and to randomly assign each respondent to one of these groups. As is constantly recommended in literature, one sample served as the primary development sample (sample 3a; n = 425), which enabled this research to evaluate the items and to arrive at a preliminary solution for the scale that seems optimal under consideration of the data. The second sub-sample (sample 3b, n = 425), the holdout sample, was then used to replicate and cross-check these findings. It is assumed that this process offers valuable information of the scale’s stability, because it enables the assessment of multiple aspects of its psychometric properties, such as its external validity and internal consistency across the split samples. For instance, if Cronbach’s alphas remain quite stable in both groups, it gives more confidence that these values are not distorted by chance (DeVellis, 2012).

This said, the development as well as the hold-out sample were both subject to the same statistical analyses. Here, a two-step approach was taken. First, exploratory factor analysis (EFA) was applied in order to investigate the scale’s dimensionality and its reliability. Together with additional reliability analyses on the item level, this procedure was used to trim the initial set of 53 items and only retain the scale’s observable indicators with reasonable statistical properties. Second, confirmatory factor analysis (CFA) was conducted to confirm the hypothesized structure of the construct and to further assess the findings of EFA. Besides the assessment of the scale’s dimensionality, CFA also enabled this research to test for aspects of convergent validity, discriminant validity among the factors, as well as internal consistency of the preliminary form of the scale (Anderson & Gerbing, 1988; Fornell & Larcker, 1981).

This research’s subsequent application of EFA prior to CFA is the result of both established recommendations for scale development (e.g., Worthington & Whittaker, 2006), as well as the regular procedures pursued in literature (e.g., Bearden et al., 2001; Shimp & Sharma, 1987). For example, Netemeyer et al. (2003) state that prior EFA is valuable, especially under conditions when item sets are large. Here, it enables researches not only to select the most appropriate scale items but also to determine the suitable number of factors. Subsequent CFA increases the researcher’s confidence in the previously identified structure when it can be successfully replicated. Also Hildebrandt & Temme (2006) acknowledge that the two-step approach is a common and useful practice. Given this, the same procedure was applied in this research.

Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science