180 



BIOLOGY OF DEATH 



The underlying idea was to get the partial or net 

 correlation between x-^ and x^, while Xq and x^ are held 

 constant. If the mortality of infancy is selective, its 

 amount should be negatively correlated to a significant 

 degree with the mortality of the next eight years when 

 the births in each district considered are made con- 

 stant and when the general health environment is made 

 constant. Under the constant conditions specified a 

 negative correlation denotes that the heavier the inf an- 



TABLE 22 

 Snow's results on selective death rate in man. 

 rural districts 



English and Prussian 



tile death rate in a cohort of births the lighter will be 

 the death rate of later years, and vice versa. The last 

 variable, x^, is the one chosen, after careful consideration 

 and many trials, to measure variation in the health envi- 

 ronment. If any year is a particularly unhealthy one — an 

 epidemic year for example — then this unhealthiness 

 should be accurately reflected in the deaths of those mem- 

 bers of the population not included in the cohort 

 under review. 



