EMPLOYMENT PSYCHOLOGY
(3) Tey = —_—
V = Test ery
mn
In this formula y represents the criterion.
The effect of doubling the length of a test having a reliability
 of .50 and correlating .40 with the criterion is to increase
 its validity to .46. If the test is increased 20 times
in length, the validity is raised to .76.
The last two formulas may be applied to the use of an
average or total score in # tests similar to x, provided these
tests are equally weighted. They apply also to » repetitions
of the same test, provided the time interval between measurements
 is long enough for memory of the preceding response
 to disappear.
It is important to know the reliability not only of test
scores, ratings, and other measurements, but also of criteria.
High correlations between measures and criteria need not
be expected unless their reliabilities are high. Foremen’s
ratings usually have low reliability in this sense, and the
investigator can scarcely expect consistently high correlations
 of test score with such a criterion.
If the reliabilities of both test and criterion have been
computed, the correlation between one set of test scores and
criterion records may be corrected for attenuation due to
errors of measurement. If 7.., is the reliability of the test
and 7,,, the reliability of the criterion, then the correlation
between test and criterion (rx) cannot be greater than
A/7rzs + Ty, Drovided similar groups are measured in a similar
 way. The size of 7., must be judged by comparison with
\/7zm1 + 73, and not with 1.00 as a maximum. The formula
for correction is then
(4) 7x (corrected for attenuation) = Te
The coefficient obtained by the use of this formula is the
correlation which may be expected between perfect measures
 of x and y (125, p. 228). The value of 7 obtained by

168