roc: Receiver Operating Characteristic (ROC)

In flr/FLCore: Core Package of FLR, Fisheries Modelling in R

Receiver Operating Characteristic (ROC)

Description

A receiver operating characteristic (ROC) curve shows the ability of a binary classifier. Here it is applied to compare two sets of values, stored as two FLQuant objects. The first is the result of aplying a logical comparison of a given state against a reference value, so it contains a binary (0, 1) label. The second, the score, contains an alternative metric that attempts to measure the absolute value of the first. The examples below compare an observation of stock status, SSB being less than a reference point, and an alternative metric, here the catch curve estimates of total mortality.

Usage

roc(label, ind, direction = c(">=", "<="))

auc(x = NULL, TPR = x$TPR, FPR = x$FPR)

Examples

data(ple4)
# OM 'reality' on stock status (fbar)
state <- fbar(ple4)[, ac(1960:2017)]
# Model estimates of F using catch curves
ind <- acc(catch.n(ple4)[, ac(1960:2017)])
# Compute TSS, returns data.frame
roc(state >= 0.22, ind)
# Needs ggplotFL
## Not run: 
ggplot(roc(state >= 0.22, ind, direction='>='), aes(x=FPR, y=TPR)) +
  geom_line() +
  geom_abline(slope=1, intercept=0, colour="red", linetype=2)

## End(Not run)
# Computes auc using the output of roc()
with(roc(state >= 0.22, ind), auc(TPR=TPR, FPR=FPR))
auc(roc(state >= 0.22, ind))

flr/FLCore documentation built on May 4, 2024, midnight