Abstract
A comparison was made of the reliability estimates given by four methods: Kuder-Richardson Formula 20 and 21 (KR20, KR21), and new short-cut estimates of KR20 given by Saupe (1961) based on previous empirical studies by Lord (1959). The tests used for the comparisons consisted of achievement examinations (Cancer Knowledge Examination and classroom tests) and tests of the attitude or personality type purporting to measure anxiety, somatization, repression, unconventionality, social desirability, dogmatism, authoritarianism, and tendency toward variety. The results indicate that the new estimates, r20* and r20′ were better estimates of KR20 than was KR21. The estimate, r20′ is on the average very close to KR20 and can be considered an accurate short-cut alternative to this coefficient. Caution should be exercised, however, in the use of the new estimates for very easy or very difficult tests since for these tests they appear to underestimate KR20.
Get full access to this article
View all access options for this article.
