Cutoff

Cutoff is a term used across disciplines to denote a boundary value that separates different states, categories, or behaviors. In statistics and data analysis, a cutoff is a threshold used to decide whether a result or observation should be considered significant or included in further analysis. Examples include p-value cutoffs (such as 0.05) for hypothesis testing or probability thresholds used to label predictions in binary classification.

In signal processing and electronics, the cutoff (often called cutoff frequency) marks the frequency at which a system’s response falls to a specified level, commonly the point where power is reduced by half (−3 dB). For filters, this defines the boundary between passband and stopband and guides design choices such as order and ripple.

In health sciences and laboratory testing, cutoff values determine whether a measurement indicates a condition or not. Reference ranges, clinical thresholds, and diagnostic cutoffs help translate numerical results into categorical interpretations such as positive/negative or above/below a clinical limit.

In computer science and machine learning, a decision cutoff converts a predicted probability into a class label. Adjusting the cutoff can affect the balance between precision and recall, and ROC curves are often used to choose an appropriate value.

Other uses include hydrology and geology, where a cutoff may describe a physical barrier that halts flow (cutoff wall) or a limit in a mined or geological seam. In everyday language, cutoff can also refer to a device that interrupts power or to a limit line that ends a process. Across fields, the core idea is a well-defined boundary used to make a decision or quantify a response.