Abstract
Statistical methods for detecting changes in longitudinal time series of psychophysiological data are limited. ANOVA and mixed models are not designed to detect the existence, timing, or duration of unknown changes in such data. Change point (CP) analysis was developed to detect distinct changes in time series data. Preliminary reports using CP analysis for fMRI data are promising. Here, we illustrate the application of CP analysis for detecting discrete changes in ambulatory, peripheral physiological data leading up to naturally occurring panic attacks (PAs). The CP method was successful in detecting cardio-respiratory changes that preceded the onset of reported PAs. Furthermore, the changes were unique to the pre-PA period, and were not detected in matched non-PA control periods. The efficacy of our CP method was further validated by detecting patterns of change that were consistent with prominent respiratory theories of panic positing a relation between aberrant respiration and panic etiology.