3 18
Abstracts of Oral Presentations
descriptors
having
retained;
significant
for interaction
retention.
Reducing
retention
treatment
effects
the panel
of only 34%,
there
reduction
beforehand,
in degrees
only
to one-quarter
resp. 9%,
significant
for these effects. The loss of information, be predicted
effects
were
This external
of the individual
allowed descriptors
which can never
is discussed
of freedom
were 34%
in relation
inherent
to the
for smaller
the products
Garmt B. Dijksterhuis ID-DLO,
Institute
for Animal
Science
drawn from quantitative
in describing
Selected
recording
research
perceived
by human resulting, recorded
of
model
for each
so-called, intensity
in
tasting
and Health,
consistencies stance.
the more
sensible,
which
perhaps
of the difficulty
is available,
A method
individual
substance.
Curves
subjects,
prototype
is investigated.
conclusions
is compared
of
curve
factors for the Time and Intensity
about
of projected
the perceived
with other methods
that over
The method
that were recently
vis. Principal
and the analysis of TI-curve
curves.
curves
intensity
to the tasted substances.
posed to model TI-curves,
is often
to interpret
However,
the possible saddle
if a clustering of consumers
thanks to one of the two methods proposed,
contour
Last but not least, external
analysis assumes that there is a valid multivariate consensus
among
trained
ences. This assumption metric
permutation
assessors about
can be checked
tests. When
consensus
can
be defined,
procedure
is proposed
pro-
Mario Bertuccioli,
Erminio Monteleone and Ella Pagliarini,
Dip. Scienze e Tecnologie (DISTAM),
Universita
50 144 Firenze,
Alimentari di Firenze
e Microbiologiche
Via Donizetti,
The analysis of hedonic
MAPPING
overall
differences
data to determine in
samples and the relationships and the sensory
attribute
acceptability between
The traditional
stimulus space to preference
France
results by the calculation ideal point models.
improvements
Mapping
techniques,
and illustrates
associated proposes
their effectiveness
with some
on real
and simulated data. The interpretation of Internal Preference Mapping, which is basically a biplot of the product consumer preference data set, can be rather tricky when the number of consumers or the number of dimensions required is large. Preference Clustering techniques are preferred for deriving homogeneous clusters of consumers. Each cluster can be summarized by its vector of product mean scores and the smaller set of these new vectors may be used for External Preference Mapping.
an ideal space.
point
For
This
that
each
if there are among
consumers’
to map individual’s
duct acceptability.
17 rue Sully, BV 1540, 2 1034 Dijon Cedex,
Preference
6 -
Italy
INRA,
some problems
to each of
41. NEW METHOD FOR PREFERENCE MAPPING
data on the sample space are relevant
identifies
weighted
preference
Pascal Schlich and Philippe Callier,
paper
multivariate
the sensory attributes.
any
The
using non para-
no valid
a new univariate
for relating
sensory
sample differ-
Curve Analysis
shapes.
40. PROGRESS IN PREFERENCE
of their shape by means of iso-
plots for instance.
curves,
To this end,
onto the prototype
using the method
time in connection
consist
but also show clear
aim of the study is to find prototype
enable
The
and for a given tasted sub-
to find underlying substances,
by means of projection The
of
These curves display large dif-
curves are projected
axes separately,
consists
as they are indicated
Time-Intensity
within a subject
for the particular
Although
enough in the sensory field. Moreover,
model,
point in preference.
AA
sensometrics
a particular
ferences between individual
calculated.
seems to be avoided in the sensory applications, because
levels over a short range of time, typi-
cally one or two minutes.
cluster
can then be
models and a weigh-
the number of response surfaces becomes small enough to
taste-intensities
subjects
consumers
on the basis of individual
to be recognized
The Netherlands
type
models can be fitted,
this selected model is significant
preference.
allow visual investigation One
analy-
to test which one is the best for each
and whether
the elliptical
P.O. Box 15, NL-7360
descriptive
but it is possible consumer,
Different
these tests are classical tools in statistics, they do not seem
and Paul H.C. Eilers,
Sensory Laboratory, Beekbergen,
DATA
regression
scores onto a sensory map of
panel.
ted mean
39. MODELLING TIME-INTENSITY USING PROTOTYPE CURVES
hedonic
sis with a trained
clustered
panels.
analysis is basically a polynomial
pro-
method that relates a
data permits
to obtain of vector
the and
finds for each subject
is positioned
of these
hedonic
to maximize
of a hierarchy method
food liking
ideals
within
the stimulus
points,
the
squared
Euclidean distances from each stimulus to the ideal point are linearly (for metric model) or monotonically (for non-metric model) expressed by the subjects. map individual’s hedonic
related to the preferences Alternatively, it is possible to data on the sample space by
using
developed
another
procedure
to obtain
only
a
model (response surface) by partial least squares modelling, called CARS0 (computer-aided response surface optimization). Using CARSO, the coefficients of the polynomial describing the surface are obtained by PLS