Type:	Package
Title:	Evaluation of Presence-Absence Models
Version:	0.2.0
Date:	2025-10-29
Maintainer:	Alberto Jiménez-Valverde <alberto.jimenez.valverde@gmail.com>
Description:	Collection of functions to evaluate presence-absence models. It comprises functions to adjust discrimination statistics for the representativeness effect through case-weighting, along with functions for visualizing the outcomes. Originally outlined in: Jiménez-Valverde (2022) The uniform AUC: dealing with the representativeness effect in presence-absence models. Methods Ecol. Evol, 13, 1224-1236.
Acknowledgments:	This work was supported by the Ministerio de Ciencia, Innovación y Universidades/Agencia Estatal de Investigación/FEDER (PID2023-147672NB-I00), and by CSIC (2024ICT253).
License:	GPL-3
Imports:	ROCR (≥ 1.0-7), stats, graphics
Depends:	R (≥ 4.3.0)
RoxygenNote:	7.3.3
Encoding:	UTF-8
NeedsCompilation:	no
Packaged:	2025-10-29 09:12:33 UTC; User
Author:	Alberto Jiménez-Valverde [aut, cre]
Repository:	CRAN
Date/Publication:	2025-10-29 10:20:02 UTC

Calculates the uniform AUC and uniform Se*

Description

This function computes the uniform AUC (uAUC) and uniform Se* (uSe*) following Jiménez-Valverde (2022).

Usage

AUCuniform(
  mat,
  rep = 100,
  by = 0.1,
  deleteBins = NULL,
  plot = FALSE,
  plot.adds = FALSE
)

Arguments

Details

This function performs the stratified weighted bootstrap to calculate the uniform AUC (uAUC) and uniform Se* (uSe*) as suggested in Jiménez-Valverde (2022). A warning message will be shown if the sample size of any bin is zero. Another warning message will be shown if the sample size of any bin is lower than 15. In such case, trimming should be considered. The AUC (non-uniform) is estimated non-parametrically (Bamber 1975). Se* is calculated by selecting the point that minimizes the absolute difference between sensitivity and specificity and by doing the mean of those values (Jiménez-Valverde 2020).

Value

A list with the following elements:

AUC: the AUC value (non-uniform), a numeric value between 0 and 1.

Se: the Se* value (non-uniform), a numeric value between 0 and 1.

bins: a table with the sample size of each bin.

suit.sim: a matrix with the bootstrapped suitability values.

sp.sim: a matrix with the bootstrapped presence-absence data.

uAUC: a numeric vector with the (uAUC) values for each replication.

uAUC.95CI: a numeric vector with the sample (uAUC) quantiles corresponding to the probabilities 0.025, 0.5 and 0.975.

uSe: a numeric vector with the (uSe*) values for each replication.

uSe.95CI: a numeric vector with the sample (uSe*) quantiles corresponding to the probabilities 0.025, 0.5 and 0.975.

References

Bamber, D. (1975). The Area above the Ordinal Dominance Graph and the Area below the Receiver Operating Characteristic Graph. J. Math. Psychol., 12, 387-415.

Jiménez-Valverde, A. (2020). Sample size for the evaluation of presence-absence models. Ecol. Indic., 114, 106289.

Jiménez-Valverde, A. (2022). The uniform AUC: dealing with the representativeness effect in presence-absence models. Methods Ecol. Evol., 13, 1224-1236.

Examples

suit<-rbeta(100, 2, 2) # Generate suitability values
random<-runif(100)
sp<-ifelse(random < suit, 1, 0) # Generate presence-absence data
result<-AUCuniform(cbind(suit, sp), plot = TRUE, plot.adds = TRUE)
result$uAUC.95CI[2] # Get the uAUC

Calculates the uniform AUC and uniform Se* by the direct weighted trapezoidal estimation method.

Description

This function computes the uniform AUC (uAUC) and uniform Se* (uSe*) using the direct weighted trapezoidal estimation method (Jiménez-Valverde 2025), instead of the stratified bootstrapping with inverse probability weighting method implemented in AUCuniform and originally proposed by Jiménez-Valverde (2022). Uniform statistics are design to account for the representativeness effect (Jiménez-Valverde 2022). This new method reduces bias and improves the coverage of confidence intervals relative to the original proposal. Additionally, the weight vector associated to each case can be customized.

Usage

AUCuniform.2(
  mat,
  by = 0.1,
  deleteBins = NULL,
  w = NULL,
  plot = FALSE,
  plot.compare = FALSE,
  plot.adds = FALSE
)

Arguments

Details

This function calculates the uniform AUC (uAUC) and uniform Se* (uSe*) using the direct weighted trapezoidal estimation method proposed by Jiménez-Valverde (2025). To compute the uniform statistics, the w parameter must be set to NULL (default). The data set is divided into bins (defined by the parameter by) based on the values of the first vector in the input matrix mat (the classification rule). Each observation is assigned a weight equal to one divided by the number of observations in the corresponding bin. Then, the uniform discrimination statistics are calculated via the direct weighted trapezoidal estimation method such that, for each threshold, the weighted true positive and false positive rates are cumulatively updated by summing the weights of the presences and absences, respectively, with that score (Jiménez-Valverde 2025). The calculation of the uniform statistics requires the classification rule (mat[,1]) to range between 0 and 1, and the value of by to divide 1 exactly. If any of this conditions are not match, the function stops. A warning message is displayed if (1) the sample size is lower than 30, (2) any bin has a sample size of zero, or (3) any bin has a sample size between 1 and 15. In the latter case, trimming should be considered using deleteBins, in which case the uniform statistics are computed excluding the selected bins. See Jiménez-Valverde (2022) for further details.

Alternatively, users may wish to downweight the importance of certain observations relative to others for reasons unrelated to the representativeness effect (Jiménez-Valverde 2025). For this purpose, the weights associated to each case can be fully customized with the w parameter (see Examples). The length of the weight vector has to be equal to dim(mat)[1].

The standard AUC (non-uniform, unweighted) is estimated non-parametrically by the trapezoidal rule, which is equivalent to the Wilcoxon-based estimation (Hanley & McNeil 1982) used in AUCuniform. Se* is calculated as in AUCuniform.

Value

A list with the following elements:

AUC: the standard AUC value (unweighted), a numeric value between 0 and 1.

Se: the standard Se* value (unweighted), a numeric value between 0 and 1.

bins: a table with the sample size of each bin (returned only if w = NULL).

uAUC: the uniform AUC value (returned only if w = NULL).

uSe: the uniform Se* value (returned only if w = NULL).

wAUC: the weighted AUC estimated with the vector w (returned only if w is not NULL).

wSe: the weighted Se* estimated with the vector w (returned only if w is not NULL).

TP: a vector with the true positive rate for every threshold.

FP: a vector with the false positive rate for every threshold.

TP.W: a vector with the weighted true positive rate for every threshold.

FP.W: a vector with the weighted false positive rate for every threshold.

References

Hanley, J. A. & McNeil, B. J. (1982). The Meaning and Use of the Area under a Receiver Operating Characteristic (ROC) Curve. Radiology., 143, 29-36.

Jiménez-Valverde, A. (2022). The uniform AUC: dealing with the representativeness effect in presence-absence models. Methods Ecol. Evol., 13, 1224-1236.

Jiménez-Valverde, A. (2025). Refining uniform discrimination metrics: towards a case-by-case weighting evaluation in species distribution models with presence-absence data. Under review.

Examples

# In this first example, a data set is simulated in such a way that the
# classification rule is well-calibrated, i.e., the observed proportion of 
# positive cases equates to the simulated probabilities of presence. Since
# the objective is to calculate the uAUC to account for the environmental 
# representativeness effect (see Jiménez-Valverde 2022), weights are 
# automatically calculated and no w vector is needed. 

n <- 1000 # Set the sample size
hs <- rbeta(n, 2, 2) # Simulated probabilities (the classification rule)
random <- runif(n)
sp <- ifelse(random < hs, 1, 0) # Observed presence–absence data

result <- AUCuniform.2(cbind(hs, sp), plot = TRUE, plot.compare = TRUE)

result$AUC  # Get the standard AUC
result$uAUC # Get the uniform AUC. Note how it is close to the reference value
            # of 0.83 since the probability values  (the classification rule)
            # are simulated to be well-calibrated (see Jiménez-Valverde 2022)

# In this second set of examples, the objective is not to calculate the  
# uniform AUC, but to assign specific weights to certain observations. These 
# examples corresponds to some of those provided in Table 1 of 
# Jiménez-Valverde (2025).

hs <- seq(1, 0.05, by = -0.05) # Generate the classification rule 
sp <- c(0, 1, 1, 0, 1, 1, 1, 1, 1, 0, 1, 0, 0, 0, 1, 1, 1, 0, 0, 0) # Observed presence–absence data

wa <- ifelse(sp == 0, 0.2, 1) # The vector of weights for each case
result.a <- AUCuniform.2(cbind(hs, sp), w = wa, plot = TRUE, plot.compare = TRUE)

result.a$AUC  # Get the standard AUC
result.a$wAUC # Get the weighted AUC. Since every case within each category of
              # sp received the same weight, the weighted AUC value equals the
              # standard AUC value

wb <- c(rep(1, 19), 0.2) # The vector of weights for each case
result.b <- AUCuniform.2(cbind(hs, sp), w = wb, plot = TRUE, plot.compare = TRUE)

result.b$wAUC # Get the weighted AUC. Since a low weight is assigned to an 
              # instance of absence associated with a low probability value, 
              # the weighted AUC is lower than the standard AUC value.
            
wc <- c(0.2, rep(1, 19)) # The vector of weights for each case
result.c <- AUCuniform.2(cbind(hs, sp), w = wc, plot = TRUE, plot.compare = TRUE)

result.c$wAUC # Get the weighted AUC. Since a low weight is assigned to an 
              # instance of absence associated with a high probability value, 
              # the weighted AUC is higher than the standard AUC value

Calculates the uniform AUC and uniform Se* by the trapezoidal method.

Description

This function computes the uniform AUC (uAUC) and uniform Se* (uSe*) using the weighted trapezoidal method instead of the weighted bootstrapping method used in AUCuniform and originally proposed in Jiménez-Valverde (2022). This procedure reduces bias and improves the coverage of confidence intervals (Jiménez-Valverde 2024). Additionally, the weights vector associated to each case can be customized. See Jiménez-Valverde (2024) for details.

Usage

AUCuniform_trap(
  mat,
  by = 0.1,
  deleteBins = NULL,
  w = NULL,
  plot = FALSE,
  plot.compare = FALSE,
  plot.adds = FALSE
)

Arguments

Details

This function calculates the uniform AUC (uAUC) and uniform Se* (uSe*) using the weighted trapezoidal method as suggested in Jiménez-Valverde (2024). A warning message will be shown if the sample size of any bin is zero. Another warning message will be shown if the sample size of any bin is lower than 15. In such case, trimming should be considered using deleteBins (Jiménez-Valverde 2022). Alternatively, the weights associated to each case can be fully customized with the w parameter (Jiménez-Valverde 2024). In this case, no warnings regarding sample size issues are raised, and deleteBins is not used. The AUC (non-uniform, unweighted) is estimated non-parametrically by the trapezoidal rule, which is equivalent to the Wilcoxon-based estimation (Hanley & McNeil 1982) used in AUCuniform. Se* is calculated as in AUCuniform.

Value

A list with the following elements:

AUC: the AUC value (non-uniform, unweighted), a numeric value between 0 and 1.

Se: the Se* value (non-uniform, unweighted), a numeric value between 0 and 1.

bins: a table with the sample size of each bin (only if w = NULL).

uAUC: the uniform AUC value (only if w = NULL).

uSe: the uniform Se* value (only if w = NULL).

wAUC: the weighted AUC estimated with the vector w

wSe: the weighted Se* estimated with the vector w

References

Hanley, J. A. & McNeil, B. J. (1982). The Meaning and Use of the Area under a Receiver Operating Characteristic (ROC) Curve. Radiology., 143, 29-36.

Jiménez-Valverde, A. (2022). The uniform AUC: dealing with the representativeness effect in presence-absence models. Methods Ecol. Evol., 13, 1224-1236.

Jiménez-Valverde, A. (2024). Improving the uniform AUC (uAUC): towards a case-by-case weighting evaluation in species distribution models. In preparation.

Examples

suit<-rbeta(100, 2, 2) #Generate suitability values
random<-runif(100)
sp<-ifelse(random < suit, 1, 0) #Generate presence-absence data
result<-AUCuniform_trap(cbind(suit, sp), plot = TRUE, plot.compare = TRUE)
result$AUC #Get the AUC
result$uAUC #Get the uAUC. Note how it is closer to the reference value of
            #0.83 since the suitability values are simulated to be
            #well-calibrated (see Jimenez-Valverde 2022).

Calibration graph

Description

A function to plot a calibration graph.

Usage

CALplot(mat, by = 0.1)

Arguments

Details

Dots for bins with 15 or more cases are shown in solid black; dots for bins with less than 15 cases are shown empty (see Jiménez-Valverde et al. 2013). This way, by plotting the calibration graph before running AUCuniform, one can get a glimpse of how reliable uAUC or uSe* can be expected to be.

Value

This function returns a calibration plot

References

Jiménez-Valverde, A., Acevedo, P., Barbosa, A. M., Lobo, J. M. & Real, R. (2013). Discrimination capacity in species distribution models depends on the representativeness of the environmental domain. Global Ecol. Biogeogr., 22, 508-516.

Examples

suit<-rbeta(100, 2, 2) # Generate suitability values
random<-runif(100)
sp<-ifelse(random < suit,1 , 0) # Generate presence-absence data
CALplot(cbind(suit, sp))

Suitability values distribution graph

Description

A function to visualize the distribution of the suitability values associated to presences, absences, and all cases together.

Usage

HSgraph(mat, breaks = 10, hist.total = TRUE)

Arguments

Details

In blue, the distribution of the suitability values associated to presences. In red, the distribution of the suitability values associated to absences. This graph helps to understand why the AUC (or Se*) is greater, equal to, or less than the uAUC (or uSe*) (see Jiménez-Valverde 2022).

Value

This function returns a multiple histogram.

References

Jiménez-Valverde, A. (2022). The uniform AUC: dealing with the representativeness effect in presence-absence models. Methods Ecol. Evol., 13, 1224-1236.

Examples

suit<-rbeta(100, 2, 2) # Generate suitability values
random<-runif(100)
sp<-ifelse(random < suit, 1 , 0) # Generate presence-absence data
HSgraph(cbind(suit, sp), breaks = 20, hist.total = TRUE)