PREDICTION OF VOCATIONAL SUCCESS 201 
A second method of treating measurements which have 
been evaluated by the method of correlation calls for the 
construction of percentile curves or tables. This process has 
been adequately explained by Thurstone (194) and Otis 
(125). The percentile graph or table gives for each test 
score the percentage of individuals who obtain that score 
or less. Instead of predicting success in terms of the cri- 
terion, this method predicts it in terms of the percentage 
of applicants whom this individual equals or excels in the 
test. Tables of deciles, quintiles, or quartiles may be cal- 
culated, when it is unnecessary to have such a fine division 
as hundredths. The percentile curve of test scores means 
nothing in the absence of a high correlation with the cri- 
terion; but if the correlation between test and criterion is 
known to be high, it may be inferred that the applicant’s 
percentile rank in the test gives also his approximate rank 
in future vocational accomplishment. 
The value of predictions made by this method, as well as 
by the method based on the regression equation, depends on 
the amount of correlation between test and criterion. 
Formula 21 may be employed in determining the validity 
of questionnaire or test items whose answers are qualitative 
and dichotomous. When this has been done, it is necessary 
to devise a scoring method for the test or questionnaire 
which shall give proper weight to the valid items. We have 
already discussed on pages 194-195 the methods of weighting 
answers to questionnaire items when group differences have 
been calculated. When formula 21 has been employed, the 
weight to be assigned to answers should be in accordance 
with values of 4 in the following formula (39): 
ThE 
bil Thre 
In this formula ¢ is the standard deviation of the frequencies 
in the categories of the x distribution, or distribution of 
answers to the items. The value of ¢ is given in the formula 
at the top of the following page (86, p. 89):