True Scores, Error, Reliability, and Unit of Analysis in Environment and Behavior Research

In: Environment and behavior: eb ; publ. in coop. with the Environmental Design Research Association, Band 26, Heft 2, S. 261-293

Abstract

The measurement of the reliability of one's instruments is an important aspect of any scientific undertaking. The use of standard methods for evaluating the reliability of environmental assessment measures has, however, recently come under attack by Richards and his colleagues. This article presents a general view of reliability and the analysis of multilevel data. The meaning of reliability and its usefulness to environmental assessment is discussed. It is shown that standard methods of reliability are useful. The split-sample correlation, which was proposed as an alternative to standard reliability co-efficients, is shown not to have the stated properties and cannot be recommended. The intraclass correlation is described and recommended as an alternative to the split-sample correlation. This article concludes with a discussion of the analysis of multilevel data.