Download as pdf or txt
Download as pdf or txt
You are on page 1of 2

Coefficient of Variation

PAT LOVIE
Volume 1, pp. 317–318

in

Encyclopedia of Statistics in Behavioral Science

ISBN-13: 978-0-470-86080-9
ISBN-10: 0-470-86080-4

Editors

Brian S. Everitt & David C. Howell

 John Wiley & Sons, Ltd, Chichester, 2005


Coefficient of Variation amount from the required route), the CV would allow
us to compare the relative variability of errors to that
of times because it does not depend on the units of
measurement.
Although the coefficient of variation (CV) is defined
The notion of the CV is generally attributed to
for both distributions and for samples, it is in the
Karl Pearson [2]. In an early article from 1896,
latter context, as a descriptive measure for data, that
which attempts to provide a theoretical framework
it is generally encountered.
for Galton’s rather informal ideas on correlation and
The CV of a sample of data is defined as the
regression, Pearson used the CV for assessing the
sample standard deviation (SD) divided by the sample
relative variability of data on variables ranging from
mean (m), that is,
stature of family members to skull dimensions of
SD ancient and modern races. Pearson pointed specifi-
CV = . (1) cally to the fact that differences in relative variability
m
indicate ‘inequality of mutual correlations’ (or, simi-
The value is sometimes expressed as a percentage. larly, of mutual regressions).
Two important characteristics of the CV are that it is A more recent application of the CV, especially
independent of both the units of measurement and the relevant to behavioral researchers working in exper-
magnitude of the data. imental fields, is as a means of assessing within
Suppose that we measure the times taken by subject variability [1].
right-handed subjects to complete a tracking task
using a joystick held in either the right hand or
the left hand. The mean and SD of times for the References
right-hand joystick (RH) group are 5 sec and 1 sec,
and those for the left-hand (LH) group are greater [1] Bland, J.M. & Altman, D.G. (1996). Statistics notes:
at 12 sec and 1.5 sec. Then, CV RH = 1/5 = 0.20, measurement error proportional to the mean, British
whereas CV LH = 1.5/12 = 0.125. Medical Journal 313, 106.
Here, we see that the relative variability is greater [2] Pearson, K. (1896). Mathematical contributions to the the-
ory of evolution. III. Regression, heredity and panmixia,
for the RH group even though the SD is only two-
Philosophical Transactions of the Royal Society of Lon-
thirds of that for the LH group. don, A 187, 253–318.
Equally well, if we had counted also the number
of errors (e.g., deviations of more than some fixed PAT LOVIE

You might also like