Karl Pearson 
33 
Whence we deduce : 
Calculated Observed 
pl = 
11166 
pi = 
1148 
h = 
27-0 
P» = 
12 
Th = 
04 
p 3 = 
0 
X 2 = 8-7 and P = '013, i.e. odds of about 76 to 1. 
Here our deviations are not as before towards multiple houses, but cancer 
apparently avoids a house where it has paid one visit! It is difficult to believe 
that the frequency should be so far below a random distribution in the case of 
multiple houses, and the matter would be still worse, if there were more cases 
relative to the houses in the City than in the Registration Area. I think we may 
safely conclude that the data for multiple houses provided by the Registrar-General 
for Ireland are probably more than 100 per cent, in error, and therefore are wholly 
worthless for the purposes for which they are apparently stated. We cannot 
suppose one visit of cancer to confer immunity on a house, and the fact that the 
multiple houses are so significantly short of the chance frequency suffices, I think, 
to discredit the data and the accuracy of the methods adopted in this Special 
Report on Cancer. 
Biometrika ix 
5 
