Measurement Model
Three methods (Cronbach’s reliability, exploratory factor analysis, and confirmatory factor analysis) were used to select and assess the final items that would be used for hypothesis testing.
Step 1. Measure reliability check. Cronbach’s measure reliability coefficient was first calculated for the items of each construct. When it reached .70, the cutoff level of reliability recommended for theory testing research (Nunnally and Bernstein 1994), the items that did not significantly contribute to the reliability were eliminated for parsimony purpose. As a result, 34 items were retained for the nine constructs: 3 each for price, store image, distribution intensity, advertising spending, and price promotions; 6 for perceived quality; 3 for brand loyalty; 6 for brand awareness/associations; and 4 for OBE. The items selected are reported in Table 1.
Step 2. Exploratory factor analysis. Exploratory factor analysis was then conducted to examine whether the items produce proposed factors and whether the individual items are loaded on their appropriate factors as intended. Factor analysis with an oblique rotation technique was conducted on all measure items, and as intended, nine distinct factors were found. Analysis with an orthogonal rotation technique also produced similar factor patterns, confirming discriminant and convergent validity of measures.
Step 3. Confirmatory factor analysis. Finally, confirmatory factor analysis was used to assess the items of the constructs more rigorously, based on the correlation matrix of the items (see Appendix A). Specifically, confirmatory factor analysis was used to detect the unidimensionality of each construct. Unidimensionality is evidence that a single trait or construct underlies a set of measures (Anderson and Gerbing 1988). This unidimensionality check updates the preceding paradigm of scale development and construct validity.