Statistical considerations in the use and analysis of single-subject designs.

Abstract

Single-subject designs measure one individual's responses to some experimental manipulation. Statistical methods exist for validly estimating the effect of an intervention on a specific individual by using data from a single-subject design. However, without strong assumptions regarding how an intervention on one individual relates to its effects on others, the results from a single-subject design provide little useful information on the general utility of the intervention. Examination of a single subject cannot verify these assumptions. Correct analysis of data from such designs allows for the possibility of correlation among the observations and the modeling of any changes over time not related to an intervention effect. When data from single-subject designs are collected, the role of assumptions in both the analysis and the generality of conclusions must be frankly acknowledged. Research often develops in stages and the single-subject design can be useful in early stages for hypothesis generation.