Quality & Quantity 35: 117–128, 2001.
© 2001 Kluwer Academic Publishers. Printed in the Netherlands.
Reliability of Coordinates Obtained by MINISSA
Concerning the Order of Presented Stimuli
, J. CAMACHO
Área de Metodología de las Ciencias del Comportamiento,
Dpto. Psicología Cognitiva, Social y
Organizacional, Facultad de Psicología, Universidad de La Laguna, 38205 La Laguna, Tenerife,
Abstract. In this work we study the stability of similarity judgements of sentences in regard to
the presentation order of stimuli through MINISSA (Michigan–Israel–Nijmegen Integrated Smallest
Space Analysis), which allows a Multidimensional Scaling with a triangular matrix of global similar-
ity indices. Two forms of the same questionnaire by inverting the order of presentation of all possible
unidirectional comparisons were applied. Results indicate that both forms are equivalent with similar
structures and a high correlation coefﬁcient between the judgements. We conclude a high stability
of the used procedure. Furthermore, in comparison with a previous study, it could be shown that the
stability of solutions obtained by MINISSA is equivalent to that obtained by INDSCAL (INdividual
Differences SCALing) which is not able to handle so many data as MINISSA.
Key words: MINISSA, Congruence Index, similarity judgement
Multidimensional Scaling (MDS) is an analytic technique for the organisation of
data that systematically classiﬁes variables considering their attributes. It refers
to a set of statistical techniques that try to measure and represent the relationship
between objects when the underlying dimensions are unknown. From a practical
point of view, MDS allows for creating an image of the information contained in
the data (Davidson, 1983; Dillon and Goldstein, 1984).
The fact that the information about the association indices can be converted
into an approximated ratio scale of interpoint distances makes MDS a powerful
technique to identify structural components of the instruments employed in psy-
chology. The variations in the sizes of the association indices, due to sampling
and measurement errors, do not distort the solution more than in other techniques,
because the errors generally affect every coefﬁcient in the matrix as a constant
(Guttman, 1959). Therefore, the solutions of MDS tend to be stable and easier to
replicate than the solutions provided by other analytical techniques.
From the point of view of methodology the development of that technique has
allowed its application to a wide range of data types and formats. In that regard, it is