Stephen J. Guastello, Marquette University, Milwaukee, WI
David E. Marra, Marquette University, Milwaukee, WI
Julian Castro, Marquette University, Milwaukee, WI
Michael Equi, Marquette University, Milwaukee, WI
Anthony F. Peressini, Marquette University, Milwaukee, WI

Abstract: This study investigated the stationarity of electrodermal time series collected in situations where turn taking in human interactions are involved. In this context, the stationarity of the time series is the extent to which a simple model can be used to fit the entire time series. The experiment involved seven participants in an emergency response simulation against one opponent. They generated 48 time series across six simulations, which were split and re-spliced to separate the team s turns and the opponent s turns. Significant differences in R2 coefficients were found for both linear and nonlinear statistical models between experimental conditions, but the difference only amounted to 3% of the accuracy of those models relative to the original data. It was thus concluded that the impact of turn taking on stationarity was a small effect at most. A comparison of synchronization coefficients for the team data, which rely on the collective accuracy of the individual time series models, indicated stronger synchronization during periods when the team was watching the opponent s actions compared to when they took their own turns. It was thus concluded, furthermore, that the common focus of attention prevailed against any non-stationarity that was introduced by turn taking.

Keywords: stationarity, time series, synchronization, emergency response, neuroergonomics