PROBABLE ERROR IN PIG-FEEDING TRIALS. 



By CHARLES CROWTHER, M.A., Ph.D. 



Department of Agriculture, University of Leeds. 



In an earlier issue of this Journal^, Robinson and Hainan have com- 

 municated the results of a statistical analysis of three sets of pig-feeding 

 experiments from which they conclude that "the probable error of one 

 animal in a pig-feeding experiment is in the region of 10 per cent, of the 

 average Kve-weight increase." 



The whole subject of the interpretation of the results of feeding 

 experiments has since been exhaustively reviewed by Mitchell and 

 Grindley^. From the results of 17 American experiments on swine, 

 involving .507 pigs divided into 49 lots of 5 to 23 pigs each, of initial 

 weights ranging from 24-7 lbs. to 354-4 lbs., and feeding periods ranging 

 from 55 to 126 days, they arrive at a figure (11| per cent.) for the 

 probable error which is substantially in agreement with the above. 

 The range of probable errors for the separate lots in all the experiments 

 quoted above is from 2| to 33 per cent, of the hve-weight increase. 



In an earlier discussion of the subject of the experimental error in 

 feeding experiments Wood and Stratton^ record an experiment with 

 four steers of similar age, weight and past history. Their live-weight 

 increases for a period of ten months, during which they received identical 

 feeding and general treatment, were so closely in agreement that the 

 probable error of any one of the four animals was only about 0-5 per 

 cent, of their average increase. Nevertheless, when put on to a fattening 

 ration they gave divergent results and "after three months' feeding 

 their probable error was as usual 13 per cent, of their average Uve- 

 weight increase." They conclude therefore that the experimental 

 error cannot be appreciably reduced by such careful selection of the 



1 This Journal, Vol. v. p. 48. 



2 University of Illinois Agric. Exp. Sta. Bulletin No. 165 (July, 1913). 

 ' This Journal, Vol. in. p. 417. 



10—2 



