Analysis and Results
The survey items that correspond to the four customercentered
relationships were tested in a four-factor confirmatory
model (details are described in the Appendix). An
acceptable fit was achieved after three items were eliminated;
this change had a negligible impact on the substantive
content of the affected dimensions, and reliability and
average variance extracted (AVE) for each dimension
were good (Table 2, Appendix). The next step was to
assess whether these four constructs were an adequate
reflection of a single higher-order construct. For this purpose,
a second-order factor structure was tested (see Figure
3), and an acceptable fit for this model was found
(Table 3). The confirmatory factor analysis (CFA) and the
second-order model provide evidence of two important
points: First, each dimension has good measurement properties
and is distinct from the other dimensions, and second,
the dimensions can be combined to form one higher exhibited homogeneity of variance violations using the generalized
least squares (GLS) procedure (Judge et al. 1985) in
the SHAZAM econometric package (White 1977), and we
estimated the remainder using ordinary least squares (OLS).
Using the approach suggested by Cohen and Cohen (1983),
we calculated the percentage of within-subjects variance
explained by time for each regression.
For owners who were less positive before the event, all
five repeated measures analyses showed significant changes
between pre- and post-event scores, with R2 values ranging
from .25 to .57. Three of the five upper-half analyses were
significant, with R2 values ranging from .04 to .08 (Table 5).
Pre- and post-event means for each group are also included
in Table 5.