143 
No. C.—STATISTICS. 
By FBANCIS G ALTON, Esq., F.B.S. 
The topics suitable to statistics are too numerous to specify ; 
they include every thing to which such phrases as u usually,” 
u seldom/’ u very often,” and the like are applicable, which vex 
the intelligent reader by their vagueness and make him impa¬ 
tient at the absence of more precise data. The principal things 
to be borne in mind in making statistical inquiries are;— 
1. That the groups with which they deal should be homo¬ 
geneous. Ex. It would be correct to inquire into the average 
height, and the frequency of different degrees of deviation from 
it, of greyhounds, and similarly as regards the speed of race¬ 
horses ; but it would be absurd to talk of these averages as 
regards dogs or horses generally, because there are many va¬ 
rieties of them differing greatly and irregularly in height and 
speed (see further remarks in par. 5). 
2. When the homogeneous groups are largely governed by 
the variation of a dominant influence, it is necessary to split 
them up into subdivisions, each referring to a short phase of 
the variation. Ex. It is correct to seek the average height of 
boys between 11 and 12, 12 and 13, and so on, but it would be 
absurd to seek that of boys generally. 
3. To select cases on a system wholly independent of the 
quality about which the inquiry is made. Ex. It would be 
correct to estimate the stature of the male adults of a nation by 
measuring individuals selected by lot; but it would be incorrect 
to take townsmen alone, and still worse those who lived in an 
unhealthy suburb of a town. It requires keen observation and 
much wariness to avoid errors due to a neglect of this caution, 
because phenomena that appear independent are often linked 
together in indirect and subtle ways. Ex. If we were to select 
the upper classes of Londoners according to the initial letter of 
their names, as printed in the Court Guide, and were to choose 
the letter Z for the purpose, we should be led utterly astray, 
as nearly all those names are foreign. 
4. As regards the requisite number of cases, a few that are 
wisely selected and accurately reported are better than very 
many that are not. Each error sacrifices several good observa¬ 
tions before it is diluted until it disappears; but if a faulty bias 
