False alarms limiting

A staggering number of papers are published each year in the literature on various candidate chemical/biological detection systems. Researchers and manufacturers make diverse claims of detection limits, sensitivity, false-alarm rates, and robustness for these systems. The committee believes that in many cases, researchers emphasize the strengths of their particular detection systems while minimizing or ignoring their flaws. This practice makes it virtually impossible to evaluate the likely performance of a detection system in real-world air transportation environments. 

Reliability. Sensors must indicate the presence of hydrogen specifically, and not provide false alarms. The sensors should have consistent reproducibility (e.g., 5% over 1 year at 2% v/v H2 in air), long-term stability and minimal drift rates. Response must not drift outside acceptable limits over that lifetime without providing an alarm. Figure 15.3 [3] shows an example comparison of the theoretical versus measured... 

In spite of recent advances in detector technology, metal detectors, mine probes, and canines remain as the most widely utilized mine detection tools. All of these methods have been utilized with some success, but all have limitations. For example, metal detectors must be operated at extremely high gain settings to obtain adequate sensitivity to detect low-metal-content mines. This is an issue because minefields are often littered with metallic objects (e.g., fragments from exploded munitions). Hence, false alarms are common, limiting the utility... 

CGM devices have the ability to alarm the patient if their blood glucose level is beyond the preset limits or even if the device projects that the patient s blood glucose level will be beyond the preset limits. The accuracy of the alarms can be measured based on the number of true alarms given, false alarms given, and alarms that should... 

Note that the first condition is the familiar Shewhart chart limits. Pattern tests can be used to augment Shewhart charts. This combination enables out-of-control behavior to be detected earlier, but the false-alarm rate is higher than that for Shewhart charts alone. 

Superimposed on this time plot are the upper and lower control limits, traditionally set at a distance of 3 times the standard error (SE) of the statistic from the center line or process mean. This controls the risk of a false alarm at a low level (a chance of 3 of 1000 if the distribution is normal). The process is said to be in a state of statistical control if the plotted points appear to occur in a random pattern and are contained within the control limits. The centerline and control limits are calculated from retrospective data from the process. 

The X chart considers only the current data value in assessing the status of the process. Run rules have been developed to include historical information such as trends in data. The run rules sensitize the chart, but they also increase the false alarm probability. The warning limits are useful in developing additional run rules in order to increase the sensitivity of Shewhart charts. The warning limits are established at 2-sigma level, which corresponds to o /2=0.02275. Hence,... 

In the absence of acceptable limits, instances of detection could be expected to result in overreaction to trivial events. If false alarms occur frequently, the consumer may become complacent and hazardous situations could be ignored. 

The amount of ATP in vegetative bacterial cells can be expected [S] to be around to 10 mg. less in spores, more in yeasts. The lower limit of the sensitivity of the technique is usually quoted to be on the order of 10 mg. This value may be even higher if background ATP has to be taken into account. The technique has two difficulties, therefore, for the detection of nonsterility— low sensitivity and the possibility of false alarms (false positives) due to extraneous ATP. 

Fig. 10. Discrimination Limits. Curve F represents the loss to society as a function of earthquake magnitude F represents the cost of avoidance (evacuation, etc.), the dashed portion simulating indirect costs associated with false alarms -- eg, mental anguish, damaged credibility, lawsuits, etc. Points of imbalance between F and F which exceed what is acceptable to society are taken as lower and upper regulatory limits, which must be matched by corresponding lower (L ) and upper (Lq) measurement limits whose difference is the Discrimination Limit (Ap). A non-zero lower limit forces an improved precision requirement in comparison to the "simple" Ln of Fig. 2.

---