Child and Adolescent Psychiatry

(singke) #1

24 Chapter 2


subsidiary hump in the tail of the main distribution (as for severe intellec-
tual disability). Second, there may be a threshold effect. In the case of be-
havioural inhibition, for example, marked inhibition as a toddler predicts
continuing shyness, whereas moderate inhibition has no such predictive
value. Finally, individuals with extreme and less extreme values on some
particular scale may differ qualitatively in other important respects. Thus,
mild intellectual disability is often associated with social disadvantage
and is not commonly associated with neurological abnormalities, whereas
severe intellectual disability is less commonly associated with social disad-
vantage and much more often associated with neurological abnormalities.
Dimensional and categorical classifications of the same phenomenon
are sometimes both valuable, but for different purposes. Blood cholesterol
provides a convenient example. There is a dose–response relationship be-
tween cholesterol level and the risk of ischaemic heart disease, with most
of the attributable risk in the population being due to the large number of
individuals with ‘high normal’ values rather than to the small number of
individuals with extremely high values. In this respect, high cholesterol is
best treated as a dimensional rather than a categorical disorder. At the
same time, individuals with extremely high levels of cholesterol are a
distinctive category from an aetiological point of view, having a Mendelian
rather than a multifactorial-polygenic disorder.


Identifying dimensions and categories
Multivariate statistical techniques now exist to help identify dimensions
and categories of disorder. Though complex in detail, the general principles
underlying factor analyses and cluster analyses are relatively easy to
understand without having to go into the mathematics (see Boxes 2.1


Box 2.1A do-it-yourself factor analysis
Look at the following list of measures that could be made on a sample of adults.
Group these measures in such a way that they correspond to two dimensions:
Height
Shoe size
Size of vocabulary
Ability to complete puzzles
Shoulder-to-elbow length
Skill at mental arithmetic

You will have had no difficulty in grouping height, shoe size and shoulder-to-
elbow length as highly correlated measures that tap an underlying dimension that
could be labelled ‘linear growth’. The remaining three measures are also highly
correlated with one another and tap the underlying dimension we normally label
‘intelligence’. The two dimensions are almost independent – you do not expect
much of a correlation between the two groups of variables, for example, between
height and size of vocabulary. Congratulations – you have carried out a factor
analysis using your intuitive knowledge of correlated and uncorrelated measures
to identify the underlying dimensions.
Free download pdf