278 
A Statii<timl Study in Cancer Death-Rates 
altering the two groups referred to above, as indicated in the diagram. As these 
groups are based on very small numbers, this step seemed justifiable. The general 
fit is slightly improved thereby*, although the form of the curve is not altered. 
In Fig. 2, the female curve is given. Here the observations are more irregular, 
which may possibly be due to the greater difficulty in obtaining correct ages in the 
/ 
— 
f 
- 
r 
i 
f 
\- 
i- 
— \ 
7 
r 
\ 
r 
\ 
Fir,. 1. Cancer Death-Eate, Males, U.S.A 
?' = ^'"(^ + 52-3713y"""('""7^52y" 
case of women. Sheppard's adjustments were used, although the result shows 
that we are not justified in supposing high contact. The total number of cases 
employed in finding the rates for the various age periods was male deaths 11,436, 
female deaths 18,039. The areas of the frequency polygons represent the number 
of cancer deaths per million of population living at each age. 
* That the theoretical start of the modified curve falls so long before birth is, perhaps, due to the 
fact that there is really not hish contact at the old age end after modification. 
