No. 636] GUINEA PIGS AND TUBERCULOSIS 



31 



moment method. Perfect positive correlation is + 1, 

 perfect negative correlation is — 1, while indicates the 

 absence. of correlation. 



Inspection of these figures reveals that all of them are 

 surprisingly low. In lots 7-9, the factor most closely 

 correlated with length of life is the rate of gain between 

 weaning (33 days) and inoculation, and even this is not 

 certainly significant in view of its probable error (+.120 

 ±.064). Larger correlations are to be expected in lots 

 10-14, owing to the greater heterogeneity in age and 

 weight, but even here the correlations are all small. Size 

 of litter, birthweight and weight at 33 days are again of 

 no value as indicators of resistance to tuberculosis. 

 There, are, however, significant correlations between 

 length of life and rate of gain (+ .223 ± .047), age 

 (+ .180 ± .048) and weight of inoculation (+ .219 ± .047). 



The degree to which variation is determined by a given 

 factor is measured by the square of the coefficient of cor- 

 relation. On this basis, only 5 per cent, of the variation 

 in length of life (i.e., 5 per cent, of its mean square de- 

 viation) is determined by the most important of the above 

 factors. The degree of determination by all of these 

 factors combined can not be found by merely adding the 

 squares of the correlations, because the factors are not 

 independent of each other. For example, rate of gain is 

 an element in determining the weight at inoculation. Any 

 effect which it has on resistance should be reflected in a 

 correlation between weight and length of life as well as 

 between rate of gain and length of life. The correlations 

 among the more important of these factors are given in 

 Table IV. 



TABLE IV 



Weight at inoculation— Rate of gain . 



-Age 



Rate of gain -Ago 



+ .780 ± .025 

 + ..'123 ± .058 



+ .553 ± .035 

 + .742 ± .022 

 -i- .670 ± .027 

 -I- .592 ± .032 



