I__. - International Military Testing Association

Figure 2’s scatterplot shows the relationship of me‘ans
to acceptor set for the 9-point scale; the R1 is .98 and the
regression line is:
y = - 26.04 + 14.71x.
Figure 1
I 2 3 4 5 6 ‘I
MEAN R4TmGS
Figure 2
MEAN RATINGS
Discussion
When we derive acceptor set sizes from the regression
equations for both the 7- ‘and g-point scales, the size of the
acceptor sets affirms what we were seeing in our data for
iridividual products. For instance, the acceptor set for a
mean of 5 on the 7-point scale corresponds to an acceptor
set size of 66% of the population, whereas a mean of 6 corresponds
with 90%. A far greater number of negative
ratings are seen for a mean of 5 as opposed to a 6: the
negative and neutral population is decreased by 25%
between means of 5 and 6.
As mentioned earlier, the 9-point scale bar been used
243
forratingfooditemsinmilitaryrationssince 1957. Senior
researchers who have spent many years in ration acceptability
feel sure they have a very good item if a rating is a
7 (“like moderately”). That is, the 7 is not a good rating
by default or relative stature of the item in the ratings list,
the feeling is that an item with a rating of 7 is very
acceptable in an absolute sense. Correspondingly, the
acceptor set picture for the g-point scale regression is: a
mean rating of 7 shows an acceptor se1 size of 77%, 7.5
shows 84%, and 8 shows 91%.
The situations described above point to how acceptor
sets can aid the defmition of product acceptability. If we
unshackle the description of product acceptance from the -
scale verbal anchors, which can make it appear that products
‘are falling somewhat short because they are not
achieving perfect scores, it may facilitate definition of
product norms that are easier to deal with both intellectually
and at gut level.
For instance, if you tell product developers that a
product is top of the line if 90% of the populace rates it
positively, the statement has an intuitive logic to it. Product
developers assume that no one product can please
lOO%ofthepopulation. Evenifthereweresuchaproduct,
it would still probably not achieve a perfect rating on any
scaled measure because there is a lot at play in the rating
game, e.g., raters tend to avoid end points on scales no
matter how they feel about a product, frames of reference
can be different among raters in regard to a product, and
even the mood the rater is in that day can affect his or her
rating.
What the norms for products should be, as defined by
the size of the acceptor set, i.e., excellent, good, average,
or poor, are to be determined. One approach might be to
determine the cumulative distribution frequencies for
acceptor sets and think in terms of percentiles. Figure 3
shows the application of this concept to the 7-point scale
data; the graph shows that an acceptor set of 45% falls in
the 25th percentile, while a set of 74% falls in the 75th
percentile. To achieve a product that scores better than
80% of all products tested, an acceptor set size of about
Figure 3
0 25 50 75 ,,m
CIJMULAINE I’ERCEh’T