Time series models of individual substance abusers.

Time series analysis is a statistical procedure appropriate for repeated observations on a single subject or unit. The goal of the analysis may be to determine the nature of the process that describes an observed behavior or to evaluate the effects of a treatment or intervention. Model identification involves specifying which of several alternative Autoregressive Integrated Moving Average (ARIMA) models best describes the series and may be used to investigate basic processes. This is illustrated by an example involving selecting the model of nicotine regulation that best represents smokers. Intervention analysis involves determining if there are any changes in level or direction for the series as a result of the intervention. Two types of applications have potential for the substance abuse area: (1) evaluation of the effects of an intervention on a single individual, and (2) evaluation of organizational-level changes (i.e., program evaluation). This is illustrated by an example that examines the effect of relaxation therapy on blood pressure. Pooled time series procedures are employed to combine the data from several different individuals or units, either by cross-sectional analysis or meta-analysis. In addition, several other issues are discussed that are critical to performing a time series analysis: selection of an appropriate computer program, alternative procedures for handling missing data, procedures for multiple observations at each occasion, and corrections for seasonal data.