2.5 Covroc right

In a coverage plot, accuracy isometrics have a slope of 1, and average recall isometrics are parallel to the ascending diagonal. In the corresponding ROC plot, average recall isometrics have a slope of 1; the accuracy isometric here has a slope of 3, corresponding to the ratio of negatives to positives in the data set.

R figure

Code

h <- 75
w <- 75 
grid.step <- 15

tpr1 <- 45
fpr1 <- 15
tpr2 <- 60
fpr2 <- 20
tpr3 <- 60
fpr3 <- 30


plot( c(0,w), c(0,h),  
     xaxs = "i",yaxs = "i",
     xaxt = 'n', yaxt = 'n',
     type = "n",
     xlab = "Negatives", ylab = "Positives")

axis(2,c(0,tpr1,tpr2,h),labels=c('0','tpr1','tpr2-3','Pos'))
axis(1,c(0,fpr1,fpr2,fpr3,w),labels=c('0','fpr1','fpr2','fpr3','Neg'))

gx <- grid.step
while (gx <= w) {
  abline(v = gx, col="gray", lty="dotted")
  gx <- gx + grid.step
}
gy <- grid.step
while (gy <= h) {
  abline(h = gy, col="gray", lty="dotted")
  gy <- gy + grid.step
}

abline(c(0,h/w),col="gray")

abline(c(30,1),lty=2) 
abline(c(0,3),lty=2)

col1 <- "blue"
points( fpr1, tpr1, col=col1, type="o")
text( fpr1, tpr1, "C1", pos=3)
abline(h=tpr1, v=fpr1, col=col1, lty="dotted")

col2 <- "red"
points( fpr2, tpr2, col=col2, type="o")
text( fpr2, tpr2, "C2", pos=3)
abline(h=tpr2, v=fpr2, col=col2, lty="dotted")

col3 <- "green"
points( fpr3, tpr3, col=col3, type="o")
text( fpr3, tpr3, "C3", pos=3)
abline(h=tpr3, v=fpr3, col=col3, lty="dotted")

Python figure

Code

import matplotlib.pyplot as plt
import numpy as np

h = 75
w = 75
grid_step = 15

tpr1 = 45
fpr1 = 15
tpr2 = 60
fpr2 = 20
tpr3 = 60
fpr3 = 30

fig, ax = plt.subplots()
ax.set_xlim(0, w)

(0.0, 75.0)

Code

ax.set_ylim(0, h)

(0.0, 75.0)

Code

ax.set_xticks([0, fpr1, fpr2, fpr3, w])
ax.set_xticklabels(['0', 'fpr1', 'fpr2', 'fpr3', 'Neg'])
ax.set_yticks([0, tpr1, tpr2, h])
ax.set_yticklabels(['0', 'tpr1', 'tpr2-3', 'Pos'])

for gx in range(grid_step, w + 1, grid_step):
    ax.axvline(x=gx, linestyle='dotted', color='gray')
for gy in range(grid_step, h + 1, grid_step):
    ax.axhline(y=gy, linestyle='dotted', color='gray')

x_diag = np.array([0, w])
y_diag = (h / w) * x_diag
ax.plot(x_diag, y_diag, color='gray')

x_diag2 = np.array([0, w])
y_diag2 = 3 * x_diag2
y_diag2_shifted = 3 * (x_diag2 - 30)
ax.plot(x_diag2, y_diag2_shifted, linestyle='dashed', color='black')
ax.plot(x_diag2, y_diag2, linestyle='dashed', color='black')

ax.plot(fpr1, tpr1, marker='o', color='blue')
ax.text(fpr1, tpr1, "C1", va='bottom', ha='center')
ax.axhline(y=tpr1, color='blue', linestyle='dotted')
ax.axvline(x=fpr1, color='blue', linestyle='dotted')

ax.plot(fpr2, tpr2, marker='o', color='red')
ax.text(fpr2, tpr2, "C2", va='bottom', ha='center')
ax.axhline(y=tpr2, color='red', linestyle='dotted')
ax.axvline(x=fpr2, color='red', linestyle='dotted')

ax.plot(fpr3, tpr3, marker='o', color='green')
ax.text(fpr3, tpr3, "C3", va='bottom', ha='center')
ax.axhline(y=tpr3, color='green', linestyle='dotted')
ax.axvline(x=fpr3, color='green', linestyle='dotted')

ax.set_xlabel("Negatives")
ax.set_ylabel("Positives")
plt.show()