Abstract
The problem of the effect of restriction of range on reliability is explored, under the assumption of classical reliability theory. Since both true score and error are truncated by selection, the standard formula based on the ratio of error variance to total variance cannot be employed for correction of range. It is also pointed out that split-halves and similar methods cannot be used to establish the reliability of the data that have been used for selection. The proper treatment of the problem is administration of an alternate form or a retest of the selected group; correction for range should be treated as a case of selection in a multivariate distribution, instead of using formulae of test theory.
Get full access to this article
View all access options for this article.