48 Cancerous Ancestry and the Incidence of Cancer in Mice. 



and non- cancerous groups. Taking the crude data, the actual percentages 

 .amongst all the offspring are : — 



Ancestry cancerous 18*2 per cent. 



Ancestry non-cancerous 8*6 „ 



Difference 9 6 „ 



these percentages being based on 340 and 222 cases respectively. When 

 a correction is made for the varying age-distributions of the two groups 

 by calculating corrected percentages based on the age-distribution of all 

 mice and reducing the numbers to the corresponding proportions per 

 thousand, the difference is merely slightly increased from 9 - 6 to 9 - 8 per 

 cent. The standard error of this difference is 2 - 96. The difference is 

 3 - 3 times the standard error, and the chance of its occurring as a mere 

 ^fluctuation of random sampling only about 1 in 1000. 



The following are the differences for the separate age-classes, with their 

 ■standard and probable errors : — 



Age. 



Differeace. 



Standard error 



Probable error* 

 = - 674o standard 

 error. 



- 9 



3-9 



4-49 



3-03 



-12 



1-3 



6-18 



4-17 



-15 



20 -4 



9-02 



6-08 



-18 



10 -5 





5-38 



-21 



25 -0 



8-59 



5-79 



-24 



5 -7 



9-51 



6-41 



24- 



2-7 



8-03 



| 5-42 



* The probable error is the fluctuation of sampling that \rill be as often exceeded as not. 



It will be seen that four of the seven differences exceed their probable 

 errors, but the only differences that do so at all considerably are those 

 for the three central age-groups. The difference between the two groups 

 is almost certainly significant, i.e. not due to mere fluctuations of sampling. 



