Miscellanea 
415 
variety for all the years it was grown. Correcting for the influence of the differentiation of 
varieties in this way I have found* 
Thus season is a far more important factor than variety in determining an individual yield. 
Illustration II. Influence of Personal Equation upon the Correlation between the Grades 
assigned to the Same Paper by a Series of Instructors. 
Stripped of the verbiage iu which it has been clothed in discussions among pedagogues, one 
of the chief problems concerning the reliability of the grades assigned in examinations resolves 
itself unto the statistical question : What is the correlation between the grades assigned to the 
same paper by difierent instructors ? 
Let g be the grade assigned to any one of m papers by any one of ii instructors, let t, , i-i be 
the "first" and "second" instructor (of a symmetrical intra-class table) passing judgment 
upon a paper, jOj, the "first" and "second" paper graded by the same instructor. Then 
from Table I of D. Starch t I deduce, by the intra-class formulae (v) — (Ix) of Biometrika, Vol. ix. 
p. 450, 
?•„. 0. = "659, !•„ a ='071. 
By using the deviation method as illustrated above, I have found 
TABLE II. 
Grades of Papers Assigned by Various Instructors. 
Instructors. 
1 
2 
3 
4 
•5 
6" 
7 
8 
9 
10 
2(^) 
1 
85 
86 
88 
85 
75 
80 
88 
87 
85 
87 
846 
2 
77 
80 
87 
80 
62 
82 
82 
87 
85 
87 
809 
3 
74 
78 
78 
75 
69 
84 
91 
83 
79 
80 
791 
Jt 
65 
65 
62 
20 
26 
60 
55 
68 
55 
50 
526 
5 
68 
82 
78 
82 
64 
88 
85 
86 
78 
80 
791 
6 
94 
87 
93 
87 
83 
77 
89 
88 
88 
89 
875 
7 
88 
90 
95 
87 
79 
85 
96 
91 
87 
89 
887 
8 
80 
84 
73 
79 
72 
83 
85 
91 
77 
76 
800 
9 
70 
70 
68 
50 
44 
65 
75 
81 
79 
79 
681 
10 
93 
92 
85 
92 
81 
83 
92 
89 
84 
85 
876 
794 
814 
807 
737 
655 
787 
838 
851 
797 
802 
7882 
Both of these results, in which an attempt was made to correct for the personal equation of the 
instructors in determining the correlation between the estimates of different instructors on the 
same paper, or to correct for the difiereiices in merit of the papers in testing the individuality of 
the instructors, are higher than the raw values given above, which are clearly spurious. Similar 
results I are obtained from Jacoby's astronomical grades §. 
* Science, loc. cit. 
t Science, N. S. Vol. xxxviii. p. 630, 1913. 
J Personally, I can attach little pedagogical significance to series as short as those of either Starch 
or Jacoby. They serve here as illustrations of method merely because I know of no more extensive 
series. 
§ Science, N. S. Vol. xxxi. p. 819, 1910. 
53—2 
