No previous visit
Previous visit without
Teltzrow et al.:Multi-Channel Consumer Perceptions
Previous visit with purchase
Table 5: Sample subgroups with corresponding sizes n and familiarity index in brackets.
No previous visit
Previous visit wi purchase
purchase Previous visit with purchase
Three groups of subjects are identified with group size N > 153 (which is the minimum sample size for structural modeling for three items for each of the four constructs [Jöreskog and Sörbom 1996b, p. 171]. These three models are calculated in the same way as the models above. Since the sample sizes for these sub-groups is smaller and we allow a variation of parameters across these nested models, the χ2 - value for the subgroups will be lower than in the overall sample, and a significant improvement of χ2 in comparison with an overall model would indicate that the construct under supervision has an effect on the relationships under consideration [Homburg and Bucerius 2005, p. 104]. This statistic is significant for all three overall χ2 – model values (see Table 6), but this is only an explorative indication because the familiarity construct itself is not included in the models, but was used to select sub-samples for the models.
For these three models, the path-coefficients between perceived reputation and trust and between perceived privacy and trust as well as the z-transformed factor score of the dependant variable trust is plotted in Figure 4. The path coefficient from perceived size to trust is not included in the diagram, because it is not significant for the three subgroups.
Table 6: path coefficients and fit indices for the three subgroups with differing familiarity
RMSEA χ2 Difference
Subgroup (familiarity index)
Path coefficient / trust factor z-score
Never visited website or store, familiarity-index 1
Purchased at store, visited site, familiarity-index 5
Purchased at website and at store, familiarity-index 6
1 0.8 0.6 0.4 0.2 0
Trust factor z-score
Path coefficient from perceived reputation on trust Path coefficient from perceived privacy on trust
Figure 4: Significant path coefficients and average trust factor scores over different familiarity indexes. The differences of the path coefficients between the groups are not significant.
First of all, the overall factor structure of the three subgroups is identical with the structure of the overall sample: perceived privacy has the strongest influence on trust, followed by perceived reputation and perceived size. Since the influence of perceived size on trust was closer to non-significance in the overall sample, it is not surprising that it did not reach significance in the tested subgroups. The fact that the overall factor structure remains constant