19 
“Were twenty-seven woups of Cree (Oxford House) and Saulteaux (Island Lake) 
Indians to be measured, the mean stature of the Cree would be found to exceed that of 
the Saulteaux by 2*5 cm., or more, in twenty-six out of the twenty-seven groups, whereas 
in one out of the twenty-seven the difference would be less than 2*5 cm.” 
. This, then, is the basis on which comparisons are to be drawn. The 
reason for saying in the above that the odds or chances are 26 to 1 is 
justified by the fact that it has been established mathematically that when 
a difference divided by the probable error of that difference is 
1 - 0 , the odds are as 1-0 to 1 * 0 ^ 
2- 0, 4-6 to 1-0 
2-3, 7-3 to 1-0 
2-5, 9-2 to 1-0 
2 - 8 , 16-0 to 1-0 
3- 0, 22-0 to 1-0 
3-1, 26-0 to 1-0 
3- 5, 54-0 to 1-0 
4- 0, 142-0 to 1-0 
5- 0, 1,341-0 to 1-0 
6 0, 19,300-0 to 1-0 
7-0, 427,000-0 to 1-0 
Clearly, then, it would in most cases appear reasonable to regard a 
ratio of 3 to 1 between a difference and its probable error as reliable 
evidence that the difference was a genuine one, because it implies odds of 
22-0 to 1-0. If the ratio be 4, 5, 6, or more, to 1, surely, when dealing 
with problems such as these, it is tantamount to proof that such differences 
would still be found to occur were we to measure the entire populations 
and not be restricted to small samples of them. 
In table III a blank space has been left where a difference, P.E. diff., 
has been found to be less than 2*0, that is to say, where odds are less than 
4-6 to 1-0. For example, we learn from table II that the Gods Lake men 
have an arm stretch of 1-2 cm. less than the Oxford House men, but we 
do not feel warranted in concluding that this mean difference of 1*2 cm. 
would be found to persist if we were enabled to measure some hundreds 
more of these men, on account of the fact that the P.E. diff. (1*18) is 
almost as great as the difference (1-2) itself. 
Arm stretch and probable error Oxford House men 183-7 ± 0-67 
Gods Lake men 182-5 ± 0-97 
1-2 
V (0-67)2 4 . (0-97)2 = sj 0-4489 -t- 0-9409 = V 1-3898 = 1-18= P.E. diff. 
Diff. /P.E. diff. = 1-2/1-18 = 1-02 = No. of times the difference is greater than 
the probable error of that difference. 
^ This is an eieerpt from table 40, “Medical Biometry and Statistics”, by Kaymond Pearl. 
