TABLE 3-1 Hypothetical Validity Relationship

 

 

Condition (C)

 

 

Absent = 0

Present = 1

Screen (S)

Neg = 0

a

b

 

Pos = 1

c

d

Sensitivity is defined as the proportion of persons with the condition (C = 1) who are flagged positive by the screen, or

(the conditional probability having a positive screen if the condition is present). Specificity is defined as the proportion of persons without the condition who are negative on the screen, or

Generally, the most effective screening tests are those with high sensitivity and specificity (e.g., .9 or so).

These two statistics are actually components of a summary measure of association derived by regressing the screen on the condition (scored as dummy variables). The regression coefficient can be calculated in this case as

or the sum of sensitivity and specificity minus one. In the case of high sensitivity and specificity measures, which are desirable for most screening tests, the regression coefficient would be quite high; for example, the regression coefficient would be .8 if both sensitivity and specificity were .9.

In the case of military enlistment standards, validity has a somewhat different meaning and therefore validity studies have taken a somewhat different approach. In the military standards context, validity means the extent to which a screen predicts a future outcome rather than an existing condition. Even a fairly weak relationship (e.g., a modest regression coefficient) between a screen and a future outcome could justify a particular standard, depending on the cost of that outcome.

For example, Table 3-2 illustrates the approximate relationship between high school education and 24-month attrition. The specificity is fairly high at .8, but the sensitivity is quite low at .4, and the overall



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement