PREDICTION BY COMBINED SCORES 2
scores. Another point not to be overlooked is the relative
dispersion of scores in the two tests. The test with the
greater standard deviation receives more weight when raw
scores are added even though the range of scores in the two
tests is the same. The method of adding raw scores should
then ordinarily be avoided.
Error due to unintentional weighting of the first sort mentioned
 in the preceding paragraph may be eliminated by
weighting to equalize the medians or means in the various
tests. If the median score in one test is five times as great
as the median score in another test, then scores in the second
test should be multiplied by five. This method, however,
does not take into account the relative dispersion of scores
in the two tests. To do this the scores in each test should
be expressed in terms of the standard deviation of scores
in that test. This step reduces to the same unit of measurement
 the scores in the different tests. To express any person’s
 score in a test in terms of the standard deviation, the
deviation of his score from the mean of the distribution is
divided by the standard deviation of the distribution, and
a plus or minus sign affixed, depending on whether his score
is above or below the mean. (If the test has a negative correlation
 with the criterion, then a plus sign is affixed before
a score which is below the mean, and vice versa.) The
probable error of the distribution should be used as the unit
measure of dispersion, if the median instead of the mean
has been calculated. The individual’s total score is the
algebraic sum of his scores in each of the separate tests, as
thus computed.
Adding percentile standings in the various tests equalizes
the weights. So, too, when decile or quintile standings are
added. A total score so obtained must be considered as a
measure of relative standing only.
The above methods assign equal weight to the various
tests. Once the matter of unintentional weighting has been
guarded against in this way, the proper weight to be given to
each test may be determined. This weight will depend both

20%