23- 



Journal of Agricultural Research 



Vol. XIX, No. s 



In 3 cases the fastest growing pig was moderately economical. 



In 2 cases the slowest-growing pig was most economical. 



In 5 cases the slowest-growing pig was least economical. 



In 5 cases the slowest-growing pig was moderately economical. 



Apparently this indicates a certain degree of correlation between the 

 characters under discussion. As a more accurate determination of cor- 

 relation between rate of gain and economy of gain the data are correlated 

 in Table XVI. For this purpose the variations, both in rate of gain and 

 economy of gain, are reduced to a percentage basis. 



XVI. — Correlation between rate of gain and economy of gain 

 Rate of gain (in percentages of the mean). 



r= —0.452 ±0.068. 



The resultant coefficient of correlation (r= —0.452 ±0.068) shows a 

 distinct negative correlation between rate of gain and economy of gain, 

 entirely disproving the apparent relation shown by Tables IX to XV. 

 The differing requirements per unit of gain are of much practical moment. 

 As has been noted, the variation in rate of gain shows a standard devia- 

 tion in percentage of 9.57 ±0.58 and an average deviation of 8.01 per 

 cent. 



POSSIBLE APPLICATION 



Pointing out applications before establishing final conclusions is as 

 dangerous as selling property without possession of title, but a considera- 

 tion of probabilities is ever in order. 



It is safe to emphasize again the danger of conclusions based on feeding 

 trials where small groups are the experimental units. If average indi- 

 vidual variations of 7 per cent are at all common, a statement in a former 

 Oregon Experiment Station bulletin* that — 



the reader should therefore hesitate at putting too much weight on differences amount, 

 ing to less than 10 per cent 



carries much weight. 



' WiTHYcoMBE, James, Potter, Ermine L.. and Samson, George R. experiments in swine feeding. 

 Oreg. Agr. Exp. Sta. Bui. 127, p. 5. 1915. 



