Statistical Methods for Causal Inference Using Single-Case Time-Series Designs

• Activity Number: 423
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Social Statistics Section
• Author(s): Bradley E Huitema*+
• Companies: Western Michigan University
• Keywords: causal inference ; single-case designs ; time-series experiments

Abstract

Alternatives to traditional randomized group experiments and clinical trials are of great interest because the practical and ethical problems associated with implementing satisfactory randomized designs are often overwhelming. A common alternative approach is to use an observational study; this is often satisfactory, but only if adequate resources are available to allow large samples and extensive covariate measurement. A very different approach is followed in behavioral psychology where small time-series experiments are the dominant approach for demonstrating causal effects. It appears that one reason for the general neglect of these designs in other fields is that adequate methods of statistical analysis are not widely known. New methods are described for several popular single-case and very small sample designs. Examples of the application of these analyses to recent applied experiments based on reversal and multiple baseline single-case designs are provided.