July 1963


Arch Gen Psychiatry. 1963;9(1):104. doi:10.1001/archpsyc.1963.01720130106016

This article is only available in the PDF format. Download the PDF to view the article, as well as its associated figures and tables.


In an over-all favorable review, published in this journal, August, 1962, of the book Clinical Research Design and Analysis in the Behavioral Sciences by Levitt, Heath stated that the book contained two errors. Levitt feels that his text was not literally incorrect, but acknowledges that his discussions might be misleading.

The first of these dealt with maximum validity. Levitt stated that the maximum validity of which a test is capable is the square root of its reliability coefficient and illustrated the point by noting that a test with a reliability of 0.64 could have a validity of 0.80. Since it is accepted in practice that validity is always lower than reliability, Heath concluded that Levitt had presented an incorrect formula. Actually if the criterion were perfectly reliable and measured the same ability as measured by the test—conditions which exist in theory only—the statement was correct. Since the book was written

First Page Preview View Large
First page PDF preview
First page PDF preview