Figure 18.5 illustrates the activity monitoring operating characteristic (AMOC) curve for PANDA over the four anthrax concentrations. Since the incubation period of the simulation is set at a minimum of 24 hours, the earliest possible detection time is shown with a dotted line at the 24-hour mark. As expected, the detection time decreases as the simulated release amount increases, since a larger release is more easily detected. In particular, at zero false positives, the detection time is approximately 132, 84, 58, and 46 hours for respective simulated release concentrations of 0.015625, 0.125, 0.5, and 1.0. The maximum widths of the 95% confidence intervals for the detection times at concentrations of 0.015625, 0.125, 0.5, and 1.0, are ±5.21, 4.00, 2.67, and 1.68 hours, respectively.
The majority of false positives with a release probability of over 50% occurred during a 10-hour period from January 20, 2002, 11:00 PM to January 21, 2002, 9:00 AM and also during a 17-hour period from midnight August 18, 2002 to August 19, 2002, 5:00 pm.
We note that the model parameters were based in part on data from the year 2000, whereas the evaluation was based on using test data from 2002. So, some false positives may have been due to a lack of synchronization between the model and the test data. When tested on data from the year 2001, there were no false alarms above the 50% level.
Was this article helpful?