:37 
SELECTION ON THE VAEIABILITY 
AND COEKELATTON OF OKDANS. 
To find the variability in stature of the selected group Ave must use (Iviii.). 
deduce: 
= 0-^3 I -287 + •200 ( y + -187 )“ + ‘387 "’•' 
We 
-- I -p ■!&/ I - — p 
(Tn! \ (Tg / cr„ 0-g 
In our particular case this gives : 
= 0-1 X -987 = 5-43. 
The actual variability in stature of French women is measured by a standard 
deviation of 5’45. Hence our selected group of men Avould be sensibly equally 
variable with French women, as far as absolute variation is concerned. 
Lastly, from (Ixi.) : 
S.1 s, 
•447 -- + -433 po 3 
CTo 
0-1 
tl3 — 
V 
—1 
3 '.-> 
•433 - -F -44/ — 0.3 
(J.. cr.T 
Avhich give in our particular case : 
r,. = X -son = -811, 
-1 
r ^3 = J X -8017 = -812. 
—1 
Such a selection, therefore, would accordingly only increase insensibly the corre¬ 
lation between stature and humerus, Avhile leaving that betAveen stature and femur 
the same. The sensible reduction of correlation betAveen stature and humerus (•SO9 
to • 771 ), Avhich is found as Ave pass from male to female, does not arise Avhen Ave 
select a group of males Avith their femur and humerus of the same length, variation, 
and correlation as those of the females. 
Illustratiox III.— Influence of a Selection of Stature in Modifying Femur 
and Humerus. 
Let us select a group of French men having the same height and variability in 
height as French Avomen, and calculate the changes Avhich Avill arise in their femur 
and humerus. 
Here the selection is giA'en by : 
/q = — 12'24, >'^' 1 = 5^45. 
W^e noAV need only the earlier formulfe of this memoir. From (1.) Ave find 
Mo = mo + -349/^, M, = m^ -f ■227/q. 
