STUDIES ON OAT BREEDING. 1 79 



1. That these coefficients vary from 3.61 percent to 19.43 

 percent. The average for the coefficients of the eleven varie- 

 ties is 11.42 percent. 



2. The Victor shows the lowest coefficient for the four 

 years. The coefficient for the Imported Scotch is only slightly 

 larger. These two varieties are by far the least variable in the 

 four year test. So far as these data go they indicate that these 

 varieties are much less afifected by seasonal variations than any 

 of the other varieties. 



3. The Irish Victor and the Prosperity show roughly twice 

 as much variation as the Imported Scotch. 



The President, Silver Mine and Senator show variation co- 

 efficients in the neighborhood of 12 percent. They are nearly 

 three times as variable as the Imported Scotch. The Lincoln 

 and Kherson proved quite variable in the four year test with 

 coefficients of over 14 percent. The Banner showed a varia- 

 bility of 15 percent of its mean, while the Regenerated Swedish 

 Select gave a coefficient of 19.43 percent or more than 4 times 

 as great as the Imported Scotch. 



Comparing the coefficients of variation for the four years 

 with those of the 191 3 plots the following points may be noted. 



1. The average coefficient for the eleven varieties is 2 per- 

 cent higher for the four year yields than for the 191 3 plots. 

 Similarily with two exceptions (the Victor and the Senator) 

 each variety shows a lower coefficient in the case of the 1913 

 plots. A difference in this direction is to be expected since, 

 when all the plots are grown in one year, they are all subject 

 to the same seasonal conditions. When different years are con- 

 sidered there are differences in the climatic conditions as well as 

 in the soil. Hence it is to be expected that there would be a 

 larger variation in the latter case. 



2. The fourth column of Table 9 shows the- difference be- 

 tween the two coefficients of variation for each variety. When 

 the coefficients for the four years production is the greater the 

 difference has been regarded as positive. From this column it 

 will be seen that while the differences are often large numeri- 

 cally, in no case is a given difference as large as three times its 

 probable error. It is customary in statistical work to regard the 

 difference between two constants as not certainly significant 

 unless the difference is at least three times its probable error. 



