Detection error tradeoff
A detection error tradeoff (DET) graph is a graphical plot of error rates for systems, plotting the false rejection rate vs. false acceptance rate. The x- and y-axes are scaled non-linearly by their (or just by logarithmic transformation), yielding tradeoff curves that are more linear than , and use most of the image area to highlight the differences of importance in the critical operating region.
Axis warping
The normal deviate mapping (or normal quantile function, or inverse normal cumulative distribution) is given by the , so that the horizontal axis is x = probit(P<sub>fa</sub>) and the vertical is y = probit(P<sub>fr</sub>), where P<sub>fa</sub> and P<sub>fr</sub> are the false-accept and false-reject rates.
The probit mapping maps probabilities from the unit interval [0,1], to the [−∞, +∞]. Since this makes the axes infinitely long, one has to confine the plot to some finite rectangle of interest.