Feb. 4, 1918 Variability of Yields of Fruit Trees and Field Trials 269 



the normal curve (see fig. 11); the mean production per tree is 186.2 ± 

 1.7 pounds, standard deviation 55-33±ii9, coefficient of variability 

 29.72 ±0.69 per cent, probable error 20.05 per cent of the mean, skewTiess 



o.ooi ±0.037. Never- 



^ r 



theless, if devoted to 

 plot experiments, a 

 difference between 

 two plots of 16 adja- 

 cent trees each, of 

 even 62.94 per cent of 

 the mean production, 

 might be due to differ- 

 ential treatment 10 

 times out of 11 and 

 due to casual varia- 

 tions of soil and trees 

 once out of 11. The 

 calculations sound a 

 note of warning 

 against drawing con- 



FiG. II.— Curve of yield of individual trees, navel orange (Antelope 

 Heights). 



elusions between such plots if the differences are less than 50 per cent of 

 the mean production of the plantation, provided we wish to have such 

 conclusions as dependable as a lo-to-i chance. 



Table VIII. — Comparison of the reliability of a plot of i6 adjacent trees with that of a 

 plot of 16 trees of four scattered units of 4 trees each. Navel oranges {A ntelope Heights) 



An example taken from the Jonathan apple orchard may well be con- 

 sidered. Suppose it is desired to know the necessary difference which 

 must exist between plots of 16 trees each to give us the reliance of a lo- 

 to-i chance that it is due to differential treatment (a) when the plots are 

 made up of adjacent trees, (b) when the plots are made up of two scattered 

 units of eight trees each, and (c) when the plots are made up of four 

 scattered units of four trees each. 



The probable errors in the above cases, a, b, and c, are ± 13.49, ±8.90, 

 and ±5.69 per cent, respectively. By proceeding as before it is found 

 that the necessary difference for a lo-to-i chance is as follows: (a) 



