|
Most
scales are designed to provide a single number that provides a
measure of performance (e.g. general mental ability), a sample
of behavior (e.g. cooperation), or a description of feelings,
attitudes, beliefs, values, opinions, or symptoms (e.g. self-esteem).
For a brief review of the four ways to construct scales (Thurstone
scales, Likert scales, Guttman scaling, and Semantic differential)
see: http://faculty.ncwc.edu/toconnor/308/308lect05.htm.
A scale
can have any number of dimensions. The Baseline Dyspnea Index
(BDI) and the dyspnea component of the Chronic Respiratory Questionnaire
(CRQ) are "multidimensional" instruments that emphasize
the kinds of activities that give rise to dyspnea in patients
as well as the effort required to complete various activities
(Mahler, Guyatt,
and Jones, 1998). The BDI, for example, includes three dimensions
of experience: magnitude of task, magnitude of effort, and functional
impairment. On the other hand, neither the BDI nor the CRQ were
designed to provide specific information about the multiple sensations
that make up dyspnea.
Our understanding
of respiratory symptoms has been advanced as much by the use of
multidimensional scales as by the use of multidimensional scaling.
Factor analysis, cluster analysis, and multidimensional scaling
are analytic techniques used to simplify data to underlying structures
or dimensions. The results of these statistical approaches have
allowed us to favor some attributes of dyspnea -- such as "effort"
and "chest tightness" -- over others. They have also
allowed for a firmer basis of theory construction in interpreting
the physiological mechanisms underlying various aspects of breathing
discomfort.
In cluster
and multidimensional scaling algorithms, individuals represent
their subjective world by rating the relative similarity (or dissimilarity)
between pairs of stimulus objects.
Question
27.1
Similarity ratings
are obtained on a scale most closely resembling:
 | Thurstone
scales |
 | Likert
scales |
 | Guttman
scaling |
 | Semantic
differential |
Cluster analysis produces
a hierarchical tree (a dendogram) showing each object at the top
and the linkages among all stimuli in a nested arrangement defined
by perceived similarity. A cluster analysis of ratings of perceived
similarity for six animals obtained in 100 healthy adults is shown
here.
| Figure
27.1: Hierarchical Clustering Solution for Dissimilarity
Judgments (n=100) Obtained between Pairs of Animals.
|
|---|
|
|
| Reprinted
by permission of Chest. http://www.chestjournal.org |
|
|