154 /304
產品受歡迎資訊與網路購物：以消費者解讀為干擾變數
154
4. Empirical Results
The data obtained for manipulation check was analyzed using MANOVA because the
values of responses were continuous and could be possibly correlated with each other. In
contrast, both the dependent variable and the independent variables are categorical, and
the dependent variable is binominal. Therefore, Generalized Linear Mixed Model
(GLMM) was used for analysis. Both independent variables were treated as fixed
variables, and products and participants were treated as random variables in the analysis.
The overall trend of participants’ responses matches expectations. Participants who
received the social condition reported of perceiving a stronger impact from other
consumers’ purchase decisions than their counterparts who received the product quality
condition (4.55 vs. 4.01,
F
(1,198) = 8.45,
MSe
= 341.74,
p
= .004). The social condition
also reported a stronger level of disappointment when the product was not purchased than
did participants in the quality condition (3.26 vs. 2.96,
F
(1,198) = 3,
MSe
= 297.08,
p
=
.084), although the difference is only marginally significant. On the other hand,
participants assigned to the quality condition perceived a stronger level of impact from
product quality than did participants assigned to the social condition, but the difference is
not significant (3.17 vs. 2.93;
F
(1,198) = 1.65,
MSe
= 346.62,
p
= .20). One possible
reason to explain the insignificant difference of quality impact is that product quality is
perceived as an essential factor for both conditions. Nonetheless, the two conditions still
responded to the three questions differently in the way expected. As for the two social
related questions, participants consistently reported a significantly higher mean in the
social condition than in the quality condition. However, the phenomenon was reserved in
the quality related question. Participants reported a higher mean in the quality condition
than in the social condition. The different response patterns between the two conditions
shows that the manipulation of social comparison and product quality is successful.
Table 1 presents the average percentages of the broad-appeal or the narrow-appeal
products being purchased by participants in the four conditions. In the condition of
congruence, consumers in both inference groups have a significantly higher percentage of
choosing a broad-appeal product over a narrow-appeal product (Quality Evaluation:
Broad-appeal (77%) > Narrow-appeal (23%);
t
(1192) = 6.16,
SD
= .193,
p
< .0001; Social
Comparison: Broad-appeal (78%) > Narrow-appeal (22%),
t
(1192) = 6.42,
SD
= .1971,
p
< .0001). That is, Hypothesis 1 is strongly supported by the empirical data in this study.