Variation and Correlation of Oats — Part II 



91 



greater amount of variability where the culms are used as units; in other 

 words, there is less variability in the averages of culm per plant than 

 in the culms making up the plants. 



Correlations 



In order to obtain a comparison of the correlations that exist when the 

 plant is used as the unit and when the separate culms of the same plant 

 are used individually as the units, seven correlation coefficients were 

 determined for each of the series. The correlation tables are shown in 

 Figs. 72 to 85. The characters correlated and the coefficients determined 

 are given in Table 2. A summary of the differences between the 

 coefficients for the same correlated characters in the two lines is given 

 in Table 3: 



TABLE 3. Summary of Differences in Correlation Coefficients for the Same 

 Characters in the Two Lines Compared 



Characters correlated 



Difference 



Difference 

 P. E. difference 



Yield and height 



Yield and number of kernels 



.001 ± .012 

 .005 ± .004 

 .007 ± .006 

 .016 ± .012 

 .080 ± .034 

 .106 ± .040 

 .130 ± .041 



0.1 

 1.2 



Yield and number of spikelets 



Yield and weight of straw . ... 



1.2 

 1 3 



Yield and average weight of kernels. 



2 4 



Average weight of kernels and height 



2.6 



Average weight of kernels and number of kernels 



3.2 







Comparing the coefficients of correlation given in Table 2, it is seen 

 that they are almost identical in the cases when yield is correlated 

 with height, number of kernels, number of spikelets, and weight of straw, 

 the differences in these cases being, respectively, .001 ± .012, .005 ± 

 .004, .007 ± .006, .016 ± .012. The difference in every case is thus 

 seen to be either less than, or practically equal to, the probable error. 

 Of these four pairs of correlation coefficients, those that are most nearly 

 identical are those for yield and height. In series 1221 this coefficient 

 is .853 ± .011, and in series 22 it is .854 ± .006 — a difference of only 

 .001 ± .012, which is but one twelfth its probable error. The correlation 



