Clinical Psychology

are believed to assess a variety of clinical skills and
achievements. Let us suppose that we use the fol-
lowing seven tests:

A IQ test
B Mathematical achievement test
C Test of spatial reasoning
D Test of analytical reasoning
E Measure of empathy
F Measure of personal adjustment
G Measure of altruism
Next, we correlate each of these tests with
every other test. This gives us a correlation matrix
in which the correlations between all possible
pairs of tests are displayed. Such a matrix, with
hypothetical correlations entered, is shown in
Table 4-3.
When we look at the correlation matrix, an
interesting pattern emerges. Measures A, B, C,
and D all show a strong positive relationship (cor-
relations range from .70 to .80). At the same time,
E, F, and G also correlate highly with one another
(correlations range from .75 to .85). But there is
virtually no relationship between the group E, F,
G and the group A, B, C, D (e.g., the correlation
between A and E is .15, between B and F is .10,
and between D and G is .12). These patterns indi-
cate that A, B, C, and D appear to be measuring a

similar underlying dimension, orfactor. Similarly,

E, F, and G belong together, suggesting a second

underlying dimension. In effect, factor analysis

does statistically with large correlation matrices

what was done here by inspection with correla-

tions from seven measures. If we had 200 mea-

sures, simple inspection would have been an

impossible task.

From the previous example, it would appear

that two factors or dimensions are involved. Let us

call them Factor 1 (derived from the correlations

among A, B, C, and D) and Factor 2 (derived

from E, F, and G). Together, these two factors

account for the significant relationships in the

matrix. Usually, these factors are then named.

This is a highly inferential phase that can occa-

sionally lead to communication problems. Some-

times the names chosen convey information

different from what was intended. However, in

our example, where Factor 1 involves A, B, C,

and D, perhaps we could choose the name Intel-

lectual Ability. Factor 2 is more difficult to name

because it includes measures of empathy, adjust-

ment, and altruism. Perhaps Healthy Altruism

would be appropriate.

Factor analysis is an especially good way of

helping organize in a coherent fashion the relation-

ships that emerge from large arrays of data. As a way

of identifying the basic elements of clinical skill (as

in the example) or those of personality, factor anal-

ysis is not the ultimate answer. After all, what

emerges from a factor analysis is determined by

the nature of the measures included in the first

place. What was not included in the sample battery

of tests to study clinical skill could hardly be

expected to turn up as factors!

Cross-Sectional Versus Longitudinal Approaches

Another way of classifying research studies is by

considering whether the studies are cross-sectional

or longitudinal in nature. Across-sectional designis

one that evaluates or compares individuals, perhaps

T A B L E 4-3 Hypothetical Correlation Matrix
for Seven Tests

Test A B C D E F G

A .70 .80 .75 .15 .20 .10

B .75 .70 .12 .10 .10

C .70 .18 .15 .11

D .12 .14 .12

E .80 .85

F .75

SOURCE: FromIntroduction to Personality, 3rd ed., by E. J. Phares. Copyright
© 1991 by HarperCollins. Reprinted by permission.

RESEARCH METHODS IN CLINICAL PSYCHOLOGY 105