CHAPTER V

CORBELATIONS  AmONG  THE  TeSTS

The original purpose in working out the correlations was to see
whether there existed a body of tests with high inter-correlations
that could be used as a composite measure of nervous and motor
control. The use of the data without reference to the ventilation
conditions under which it was collected is permissible since it was
not possible to trace any ventilation effect in the records. Accord¬
ingly for every test the data for any one individual were arranged
in two equal groups to permit of correction for attenuation due to
chance errors, which do not balance out in the case of correlations
as in the case of averages, but cause the coefficient to approximate
zero. This division into two groups was accomplished by adding
alternate tests, putting the trials numbered 1 into Group I, the
trials numbered 2 into Group II. The following table shows the
chronological position of tests added to form Group I and Group II
for every individual in every test, the procedure varying somewhat
according to whether there were two or three tests a day.

A.M.          P.M.                 A.M.         P.M.j^         P.M.n

Monday...............1               2                      12                  1

Tuesday...............2               1                      2                 1                   2

Wednesday............1               2                      1                 2                   1

Thursday..............2               1                      2                 12

Friday.................1               2                      1                 2                   1

Monday...............2               1                      2                12

Tuesday...............1               2                      12                  1

Wednesday............2               1                      2                1                  2

Thursday..............1               2                      12                  1

Friday.................2               1                      2                12

On the few occasions when a test in one group was missing, the cor¬
responding test was dropped from the second group. An average
was computed for each individual in both groups of every test and
these averages treated as the original measures to which was appUed
the Pearson correlation formula:

XX'Y

and the Spearman correction formula used by Thorndike in the
shortened form:

56