Do We Have Enough? The Accuracy of Incomplete Data From Small Groups.

This article investigates the accuracy of incomplete data from research on small work groups. One area given little attention in organizational research is the number of responses needed from a unit in order to construct composition and compilation models accurately. Within the organizational literature, two approaches seem to have been used. Some researchers have decided to adopt a minimum proportion method, while others have used a minimum number method. By examining sampling theory more carefully, we can make a more informed decision. But first, it is worth considering the types of variable we may want to measure in groups based on individual responses. These essentially fall into two categories, composition models and compilation models. In summary, the recommendation to researchers who are working with incomplete data from small groups would be to take care over deciding whether or not to include a group for analysis: do not blindly use a fixed criterion for inclusion based on either response rate or minimum number of responses, but choose groups on the basis of how accurate their estimates are likely to be.