roc.plot: Make a Receiver Operating Characteristic plot

In davidavdav/ROC: Compute structures to compute ROC and DET plots and metrics for 2-class classifiers.

Make a Receiver Operating Characteristic plot

Description

These are basic plotting routine of objecs of type roc, ROC and DET plots.

Usage

 plot(r, nr=1, chull=TRUE, type = ifelse(nrow(r) > 15, "l", "b"), 
 	 traditional = F, xlim = c(0, 1), ylim = c(0, 1), 
	 xlab = NULL, ylab = NULL, lwd = 2, ...)
 roc.plot(x, ...)
 det.plot(x, nr=1, lty=1, col=nr,  xmin=0.1, xmax=50, ymin=0.1, ymax=50,
         xlab="false alarm probability (%)", ylab="miss probability (%)",
         chull=F, Ninterp=10, ...)

Arguments

r	An object of class roc
nr	The number (color) for this plot. nr=1 starts a new plot.
chull	Plots the ROC convex hull instead of the ROC.
type	The plotting symbol/line type to use.
traditional	If TRUE, the true positive rate rather than it complement, the false negative rate, is drawn.
xlim, ylim	Plotting limits
xlab, ylab	x and y labels
lwd	plotting line width
x	An object of class roc, or an object that can be coerced into that.
lty	Line type for this plot
col	Color for this plot
xmin	Minimum of the False Alarm rate, in percent
xmax	Maximum of the False Alarm rate, in percent
ymin	Minimum of the Miss rate, in percent
ymax	Maximum of the Miss rate, in percent
...	Additional arguments passed to plot.

Details

plot.roc draws a curve showing the trade-of betwee false positives (x-axis) and false negatives (y-axis) by varying an implicit threshold for the score. Better discrimination between tatrgets and non-targets leads to a curve more to the lower left, i.e., fewer errors in general.

The default is to draw the trade-off between errors. Others prefer to see the trade-off between true and false positives, for this the argument traditional=TRUE can be used.

A heuristic is made about the plotting as a continuous or discrete curve with points, if the number of points on the ROC (or convex hull) is more than 15 lines are used, otherwise points and lines.

det.plot shows the same information as a ROC plot, but on a warped probability scale. The worping function is qnorm, the inverse of the cumulative normal distribution. If the underlying target and non-targets score distributions are Gaussian, this results in a straight line in the DET plot. Often, straight lines are observed in the DET plot, suggesting that the scores can be transformed such that the scores are more-or-less normally distributed.

A special value of x=NULL causes an empty DET frame to be plotted, with just a grid and labeled axes.

Multiple DET curves can be made in the same plot by setting either nr>1 or lty>1.

DET plots are traditionally plotted with False Alarms (or False Positives) along the X-axis and Misses (of False Negatives) along the y-axis. Although it is possible to change the range of the axes this is discouraged; the general position of the curve within the plot gives the experienced DET plot reader an immediate feeling for the discriminability of the detector.

In psychology the measurement of miss and false alarm rates can be converted to a quantity d' (d-prime), which in a DET plot can be found by moving from the measured point (false alarm rat, miss rate) in the graph along a line of angle -45^\circ until the diagonal is crossed. The values of d' vary from 0 in the upper-right to about 6 in the lower left of the graph.

Value

A data frame with the summary statistics of the ROC.

Author(s)

David A. van Leeuwen

References

Alvin Martin et al, “The DET Curve in Assessment of Detection Task Performance,” Proc. Interspeech, 1895–1898 (1997).

Examples

z <- cst.tnt(rnorm(100, 1), rnorm(1000, 0))
plot(roc(z))
det.plot(z)

davidavdav/ROC documentation built on Sept. 8, 2023, 2:39 p.m.