ABSTRACT
The consideration of time is particularly important in single-subject behavioral studies, for which the time dimension is frequently the only source of data variation. Numerous researchers have demonstrated or suggested that serial dependency is a common phenomenon in studies in which behavioral observation data are used. When the purpose of an analysis is to describe a behavior, as in an ex post facto study, or to assess the effects of a particular behavioral intervention method, as in a quasi-experimental design, the existence of serial dependency renders most of the conventional statistical methods inappropriate. The chapter presents two statistical indicators of serial dependency: the autocorrelation function and Young's C statistic. The autocorrelation function (ACF) is fundamentally a set of Pearson's product-moment correlation coefficients. There are a number of methods to test for the statistical significance of an autocorrelation. The most common method is the Bartlett test.
