Test length and validity revisited

Author:
Publisher website
Google Scholar
Abstract
Results from three different subject samples extend earlier findings that lengthening a scale beyond some point can actually weaken its validity. A near‐optimal algorithm selected the most valid aggregate of items from a common pool. Findings were then cross validated in a second sample. From this procedure emerged fairly short scales with acceptable cross validities, but only if the item pool had been prescreened for content saturation. Under these circumstances, even extremely short scales of two to four items each, which had survived double cross validation, suffered hardly any loss of cross validity. In a third sample they outperformed standard scales eight times as long. © 1997 John Wiley & Sons, Ltd.

This publication has 12 references indexed in Scilit: