No. 636] GUINEA PIGS AND TUBERCULOSIS 



35 



and a slightly greater direct effect of rate of gain, pre- 

 sumably as a measure of the condition of health. 



.9884 ill lot« 7-fl 



method of analysis applies strictly only where the effects of the factors are 

 not perfectly met in the present case. 



The facts brought out by the method of correlation are 

 presented graphically in Charts 2, 3, 4 and 5. These 

 charts show the decline in numbers, on the basis of 100 

 per cent, alive on the 15th day after inoculation among 

 groups of guinea pigs of lots 10-14, classified by weight 

 at 33 days, weight at inoculation, and rate of gain. 



As a result of the foregoing considerations, it must be 

 concluded that the apparent condition of a guinea pig at 

 the time of inoculation and a knowledge of its past his- 

 tory give exceedingly little indication as to its probable 

 length of life after inoculation. Over 98 per cent, of the 

 variation in length of life iii lot< 7 9 and over 93 per 

 cent, in lots 10-14 is caused by lad. )rs otli. r tlian those 

 discussed. This leads us to a consideration of hered- 



