Multidimensional Scaling

1
Multidimensional Scaling
Multidimensional scaling (MDS) can be considered
to be an alternative to factor analysis. In
general, the goal of the analysis is to detect
meaningful underlying dimensions that allow the
researcher to explain observed similarities or
dissimilarities (distances) between the
investigated objects. In factor analysis, the
similarities between objects (e.g., variables)
are expressed in the correlation matrix. With MDS
one may analyse any kind of similarity or
dissimilarity matrix, in addition to correlation
matrices.
2
Multidimensional Scaling
Collecting and analyzing data in multidimensional
scaling experiments A guide for psychologists
using SPSS G. Giguère Tutorials in Quantitative
Methods for Psychology 2007, Vol. 2(1), pp.
26-37. Paper A Review of Multidimensional Scaling
(MDS) and its Utility in Various Psychological
Domains. N. Jaworska and A. Chupetlovska-Anastasov
a Tutorials in Quantitative Methods for
Psychology 2009, Vol. 5(1), pp. 1-10. Paper
3
Multidimensional Scaling
The data are from an experiment in which subjects
were asked to look at a screen, which had two
circular opaque glass windows. These windows were
lit from two projectors behind the screen.
Different colour filters could be inserted in the
projectors. Fourteen colour filters were used.
Each stimulus was combined with each other
stimulus in a random order. The subjects were
then asked to rate the degree of qualitative
similarity between each pair of colour filters
on a 5-point scale. The similarity matrix is
provided.
4
Multidimensional Scaling
Analyze gt Scale gt Multidimensional Scaling
(PROXSCAL)
5
Multidimensional Scaling
A single matrix of proximities has been provided.
6
Multidimensional Scaling
Choose define and select all proximities.
7
Multidimensional Scaling
Choose model, ordinal similarities.
8
Multidimensional Scaling
Choose Torgerson as an initial configuration in
options
9
Multidimensional Scaling
The first analysis may now be performed.
10
Multidimensional Scaling
Strictly you should employ numerous random
restarts to ensure a true minimum has been
located. The current solution suggests that only
one dimension should suffice.
11
Multidimensional Scaling
Restrict the model to one dimension.
12
Multidimensional Scaling
Restrict the model to one dimension.