Abstract
The desirability of performing tests of significance of measures of item validity (in item analyses) has been noted in a number of texts (Magnuson, 1966; Guilford and Fruchter, 1973; Anastasi, 1968). However the problem of determining the significance level (α) which should be used in these statistical tests, in order to control the risk of occurrence of an excessive number of Type I Errors has, apparently, received little attention. The present paper addresses itself to this problem, limiting itself to the case in which the tests of significance of the item validity measures are independent.

This publication has 0 references indexed in Scilit: