Contributors to a signal from an artificial contrast

Data from a process or system is often monitored in order to detect unusual events and this task is required in many disciplines. A decision rule can be learned to detect anomalies from the normal operating environment when neither the normal operations nor the anomalies to be detected are pre-specified. This is accomplished through artificial data that transforms the problem to one of supervised learning. However, when a large collection of variables are monitored, not all react to the anomaly detected by the decision rule. It is important to interrogate a signal to determine the variables that are most relevant to or most contribute to the signal in order to improve and facilitate the actions to signal. Metrics are presented that can be used determine contributors to a signal developed through an artificial contrast that are conceptually simple. The metrics are shown to be related to traditional tools for normally distributed data and their efficacy is shown on simulated and actual data.