For robust detection performance, alerting algorithms for biosurveillance require input data free of trends, day-of-week effects, and other systematic behavior. Time series forecasting methods may be used to remove this behavior by subtracting forecasts from observations to form residuals for...