Detectiecurves

Detectiecurves are curves that describe how the probability of detecting a signal varies with a parameter such as stimulus intensity, signal strength, or signal-to-noise ratio. They are used across disciplines, including psychology, neuroscience, and engineering, to summarize detection performance and to estimate perceptual or system thresholds.

In psychophysics, detection curves plot the likelihood of detection or a correct response as a function of stimulus intensity. The resulting relationship is often sigmoidal and can be described by psychometric functions such as the logistic or the cumulative Gaussian. Key concepts include the detection threshold (the stimulus level associated with a chosen performance criterion, for example 50% detection) and the slope (which reflects sensitivity). Experiments may use adaptive methods like staircase procedures to efficiently converge on the threshold.

Signal detection theory provides an additional framework by separating sensitivity from decision bias. Detection curves can be combined with measures of false alarms and hits, leading to derived quantities such as d' (d-prime) and decision criteria. Receiver operating characteristic (ROC) curves, which plot hit rate against false-alarm rate, are related tools that describe performance across different biases.

In engineering and applied sensing, detection curves evaluate sensors and detectors (for radar, medical imaging, or environmental monitoring). Here the curves often plot detection probability against performance metrics such as SNR or amplitude, helping to specify operating ranges and reliability.

Estimating a detection curve typically involves fitting a model to observed responses, using methods like maximum likelihood estimation, and validating the fit with resampling or cross-validation. Limitations include dependence on experimental design, observer criteria, and assumptions of the chosen model.