282 A Statistical Study in Cancer Death-Rates 
Cancer Death-Rates. 
State 
Crude 
Corrected 
Maine ... 
837 
692 
New Hampshire 
668 
568 
Vermont 
818 
647 
Massachusetts 
664 
648 
Rhode Island 
623 
590 
New York 
612 
587 
Connecticut 
611 
563 
New Jersey 
504 
472 
Michigan 
580 
498 
It will be noted that this correction although lowering all the rates does not 
materially disturb their order. 
The correction factor as used for the various States is based only on the age 
distribution of the whole State without reference to the proportions in the urban 
and rural areas. The necessary figures to make this latter correction were un- 
obtainable. 
Throughout the paper the correction factors are based on the census of 1900 
while the death-rates of the cities are those of the five-year group 1900 — 1904 ; it 
being assumed that the age distribution has kept fairly uniform, or at least has 
not altered sufficiently to affect materially the conclusions arrived at. In the case 
of the States the 1906 death-rate is compared with the insanity figures for 1904. 
This was unavoidable as no others were obtainable. I do not suppose however 
that the rates changed so greatly in two years that all value from these com- 
parisons is lost. 
In calculating the coefficients of correlation and their probable errors a doubt 
arose as to whether it was right to give every city the same weight. Should a city 
with three million inhabitants be given greater weight than one of only 100,000 ? 
If it were proved that any factor, such for instance as a meteorological condition, 
determined the incidence of cancer, it is clear that weighting for size would be 
wrong. Again if overcrowding was a marked factor it would be unjustifiable 
to weight for size. To take 100,000 inhabitants as a unit was quite arbitrary; 
I have therefore regarded each city or State as a unit. This reduces to a minimum 
the value of the denominator in the probable error, which is therefore of maxi- 
mum size. 
When first approaching this problem I was of opinion that in spite of the trend 
of modern research it might yet be demonstrated that cancer was of infectious 
origin, either bacterial or pi'otozoic in nature. Many infectious diseases have 
a very marked seasonal incidence, and are in certain cases closely correlated with 
various meteorological conditions. That this is so in respect to enteric fever in the 
Transvaal I have shown elsewhere, and it is I think reasonable to suppose that if 
