BIOMETRY 



69 



ample, if I knew the height of each member of a college 

 class I could calculate the absolute average height of the 

 class without any possible inaccuracy, if the arithmetical 

 operations were free from mistakes. But if I want to know 

 the average height of students in the entire college and have 

 only the measurements of a particular class on which to base 

 an opinion, it is obvious that my conclusion is possibly 

 erroneous. Perhaps I have not a fair sample of the students 

 of the college as regards height. Obviously the larger my 

 class the less probable is any error in my conclusion. If my 

 class included half or more than half of all the men in the 

 University (unselected as to size) the probability of an error 

 through random sampling would be small; and if it included 

 all men in the University, the probability of error would 

 disappear. 



TABLE 4 



Correlation Table used as a Measure of Heredity. The Char.4cter 



Studied is the Relative Amount op White in the " Hooded " 



Pattern of Piebald Rats, r = 0.233. 



From Castle and Phillips, Table 11. 



What statisticians call the probable error is a pair of values 

 one larger than the calculated value, one smaller, the 

 chances being even that the true value lies inside or outside 

 the limits of these values. 



To understand the significance of this statement, consider 

 for a moment the normal curve or curve of error (Fig. 11). 

 On either side of its mean and mode (3/) we may draw a line 



