Experimental procedure
Rationale for using adaptive conjoint analysis
Instead of directly asking a subject how important an attribute/
level combination is, we apply a decompositional approach. To
increase ecologic validity of the results, participants are asked to
evaluate full product profiles in a CA setting. Because of the large
number of attributes (three intrinsic and three extrinsic character-
istics, 16 combinations in each category), which allows 256 combi-
nations, we applied an adaptive, computer-based variant of CA. By
combining compositional and decompositional methods, ACA is
capable to reliably estimate individual preferences for a large set
of attributes while using a manageable number of choice tasks
(Orme & Johnson, 1996). For several reasons, we omitted the com-
positional part by setting equal importance for all attributes and
their levels: (1) Evaluation of single attributes in the compositional
part is risky because respondents may guess the purpose of the
study. (2) Attribute evaluation in the compositional part is gener-
ally based on self-report questionnaires. This would particularly
950
K. Hoppert et al./Appetite 59 (2012) 949–955
bias the evaluation of intrinsic attributes and, in turn, impair the
estimation of the utility value of an attribute in the decomposition-
al part of ACA. (3) The compositional part was omitted to hide at
which attributes our study was focusing on. In our study setting,
the participants have to make a trade-off between full product
profiles so that they are not able to overrate the importance of
single attributes. This, in turn, reduces the risk of socially desirable
responses and minimises the risk of overstressing the relevance of
a particular attribute (Carroll & Green, 1995; Louviere & Islam,
2008).
General schedule of the process
Upon arrival in our sensory laboratory (22 ± 1 ?C), each partici-
pant was seated at a central table and instructed that the task was
to carry out nine preference tests with pairs of vanilla yoghurt and
some additional tests such as answering a questionnaire. The num-
ber of paired comparisons was set to nine to ensure a reasonable
number of tests without overstressing the capacity of the partici-
pants. Personal information concerning age, gender, body height
and weight, shopping behaviour and yoghurt consumption was
gathered; the other questions were not related to the objective of
the study. To distract the participants in the breaks between
servings, they were allowed to solve a mechanical or a cross-word
puzzle. The participants gave written consent beforehand, and they
were instructed to ring a bell once a particular task was finished. A
stainless steel spoon was provided, as was tap water for
neutralisation.
Experimental procedure
Rationale for using adaptive conjoint analysis
Instead of directly asking a subject how important an attribute/
level combination is, we apply a decompositional approach. To
increase ecologic validity of the results, participants are asked to
evaluate full product profiles in a CA setting. Because of the large
number of attributes (three intrinsic and three extrinsic character-
istics, 16 combinations in each category), which allows 256 combi-
nations, we applied an adaptive, computer-based variant of CA. By
combining compositional and decompositional methods, ACA is
capable to reliably estimate individual preferences for a large set
of attributes while using a manageable number of choice tasks
(Orme & Johnson, 1996). For several reasons, we omitted the com-
positional part by setting equal importance for all attributes and
their levels: (1) Evaluation of single attributes in the compositional
part is risky because respondents may guess the purpose of the
study. (2) Attribute evaluation in the compositional part is gener-
ally based on self-report questionnaires. This would particularly
950
K. Hoppert et al./Appetite 59 (2012) 949–955
bias the evaluation of intrinsic attributes and, in turn, impair the
estimation of the utility value of an attribute in the decomposition-
al part of ACA. (3) The compositional part was omitted to hide at
which attributes our study was focusing on. In our study setting,
the participants have to make a trade-off between full product
profiles so that they are not able to overrate the importance of
single attributes. This, in turn, reduces the risk of socially desirable
responses and minimises the risk of overstressing the relevance of
a particular attribute (Carroll & Green, 1995; Louviere & Islam,
2008).
General schedule of the process
Upon arrival in our sensory laboratory (22 ± 1 ?C), each partici-
pant was seated at a central table and instructed that the task was
to carry out nine preference tests with pairs of vanilla yoghurt and
some additional tests such as answering a questionnaire. The num-
ber of paired comparisons was set to nine to ensure a reasonable
number of tests without overstressing the capacity of the partici-
pants. Personal information concerning age, gender, body height
and weight, shopping behaviour and yoghurt consumption was
gathered; the other questions were not related to the objective of
the study. To distract the participants in the breaks between
servings, they were allowed to solve a mechanical or a cross-word
puzzle. The participants gave written consent beforehand, and they
were instructed to ring a bell once a particular task was finished. A
stainless steel spoon was provided, as was tap water for
neutralisation.
การแปล กรุณารอสักครู่..
