th.indep: Spatial threshold-independent accuracy measures

Description

Calculates spatially corrected, threshold-independent metrics for an observational data set and model predictions (AUC, ROC, max-TSS)

Usage

th.indep(data, coord, spatial = TRUE, plot.ROC = FALSE, customize_plot = NULL)

Arguments

data

A data frame or matrix with two columns. The first column should contain actual presence/absence data (binary, 0 or 1) and the second column should contain model predictions of probability of occurrence (numeric, between 0 and 1).

coord

A data frame or matrix with two columns containing x,y coordinates for each actual and predicted value. Coordinates must be integer and consecutively numbered.

spatial

A logical value indicating whether spatial corrected indices (rather than classical indices) should be computed.

plot.ROC

A logical indicating whether the ROC should be plotted. NOW DEPRECATED.

customize_plot

Additional plotting parameters passed to ggplot. NOW DEPRECATED.

Value

A list with the following components:

AUC: Area under curve
opt.thresh: optimal threshold for maximum TSS value
TSS: Maximum TSS value
sensitivity: Sensitivity
Specificity: Specificity
AUC.plot: A ggplot object

References

Carl G, Kuehn I (2017) Spind: a package for computing spatially corrected accuracy measures. Ecography 40: 675-682. DOI: 10.1111/ecog.02593

Examples

Run this code

# NOT RUN {
data(hook)
data <- hook[ ,1:2]
coord <- hook[ ,3:4]
si2 <- th.indep(data, coord, spatial = TRUE)
si2$AUC
si2$TSS
si2$opt.thresh
si2$plot

# }