292 A Statistical Study in Cancer Death-Rates 
and newspapers printed in the various States to the rate per head of population, 
and correlating with cancer death-rate ; the value thus found is : 
/3 = -6182 ± -1076. 
I record these observations for any interest that they may have, but do not think 
that the deductions should be pushed too far. 
Another aspect of the question that requires investigation is the death-rate 
among the various occupations. No reliable insanity figures can be obtained that 
correspond to the grouping for occupations as given in Vital Statistics, Vol. I., 
U.S.A. Census, 1900. In fact the relation of insanity to occupation has not been 
worked out for the United States. As the Editor of the Report above referred to 
says "On the whole, the question of the relation of insanity to occupation is one 
that still awaits systematic treatment. The available means of enquiry do not 
suffice even for a perfect classification of occupations, much less for obtaining data 
which permit exact comparisons with the total number of the insane drawn from 
any given occupation with the total number of the same age, &c., engaged in that 
particular occupation." 
Considerable difficulty arises in correlating cancer and diabetes in respect 
to occupations owing to the large value of the Coefficient of Variability of the 
correction factor for age distributions, in the various occupations. An inspection 
of these factors will show that an appreciable positive correlation may be expected 
from this cause, on the other hand the value of p' will contain a spurious corre- 
lation of unknown size. Taking into consideration these facts and also that the 
negative sign of the correlation as found for the eight groups is unchanged even 
when uncorrected deaths are used, it will probably be safe to conclude that 
the correlation between cancer and diabetes death-rates with occupation groups 
is negative. 
Occupations. Cancer and Diabetes. 
8 Inclusive Groups p = - -0045 + -2375, p = - -6719 ± -1308. 
28 Sub-groups p = -2866 ± -1170, p= - -4832 + -0977. 
At first sight this may seem unexpected, but I do not think it is really surprising. 
The rates of both diseases vary considerably in the different occupations. It must 
be remembered that certain occupations consist of highly selected individuals, and 
thus the unstable in certain respects may have been weeded out. Whereas a 
similar condition imposed on a city or district may lead to an increase in both 
rates, conditions of work &c. may determine which form the manifestation of 
instability shall take. For instance occupations associated with a sedentary life 
probably predispose to diabetes. 
The use of an analogy will better explain my meaning. Tlie death-rates from 
Typhoid Fever and Bullets during time of war would be much increased, and both 
in relation to time and place would show positive correlation with war. But if 
