140 



Prohabk' Error in riy-Jecdiii;/ Trials 



In order to facilitate comparison the data have been condensed in 

 Table II into 3-\veekly periods, for which the averages of each group 

 are set out separately. 



The results present two features of special interest in the present 

 connection, viz. the extremely low degree of variability between the 

 individual records and the absence of any marked tendency for the 

 probable error (relative to increase) to fall as the feeding progressed. 



Table II. 



Weeks 



of 



experiment 



4-6 



7- 9 

 10-12 

 13-15 

 16-18 

 19-21 

 22-24 



Average 1 

 (21 weeks) ) 



Average ^ 

 for 1 8 weeks 

 (excluding > 

 4th-(ith 

 weeks) ) 



Average gain in weight 

 for 3 weeks 



• Probable error of one pig expressed 

 as percentage of average gain 



Group A 



lb. 

 ll-8± 

 17-7± 

 23-5± 

 16-8± 

 23-6± 

 24-8± 

 30-4± 



148-6±2-l 



136-8±l-8 



Group B 



lb. 

 ll-8± 

 15-4± 

 22-l± 

 19-0± 

 23-0± 

 27-9i- 

 30-6-fc 



149-0±3-7* 



138-0i2-6 



All pigs 



% 



7-6t 



7-9 



5-8 



7-9 



5-4 



8-4 



5-7 



3-7t 



3-4 



* 4 pigs only. 



t 9 pigs only. 



It will be noted from Table I that, with one exception towards the 

 end of the feeding period, the probable error of one pig never exceeded 

 21 per cent, of the average weekly increase, whilst for 3-weekIy periods 

 (Table II) the maximum (all pigs) was only 8-4 per cent. For the whole 

 period (9 pigs for 21 weeks or 10 pigs for 18 weeks) the probable error 

 of one pig was only about 3J per cent, of the average increase — as 

 compared with the general average for pigs of 10-12 per cent, quoted 

 by Robinson and Hainan, and by Mitchell and Grindlcy. If the eight 

 male animals alone be considered the probable error is only 2-9 per 

 cent. 



' The relative constancy of the probable error throughout the course 

 of the feeding, though not without precedent, is more surprising and 



