optiThresh: Optimize threshold for model evaluation.

Description

The 'optiThresh' function calculates optimal thresholds for a number of model evaluation measures (see threshMeasures). Optimization is given for each measure, and/or for all measures according to particular criteria (e.g. Jimenez-Valverde & Lobo 2007; Liu et al. 2005; Nenzen & Araujo 2011). Results are given numerically and in plots.

Usage

optiThresh(model = NULL, obs = NULL, pred = NULL, interval = 0.01,
measures = modEvAmethods("threshMeasures"),  
optimize = modEvAmethods("optiThresh"), simplif = FALSE, 
plot = TRUE, sep.plots = FALSE, xlab = "Threshold", ...)

Arguments

model

a model object of class "glm".

obs

a vector of observed presences (1) and absences (0) or another binary response variable. This argument is ignored if 'model' is provided.

pred

a vector with the corresponding predicted values of presence probability, habitat suitability, environmental favourability or alike. This argument is ignored if 'model' is provided.

interval

numeric value between 0 and 1 indicating the interval between the thresholds at which to calculate the evaluation measures. Defaults to 0.01.

measures

character vector indicating the names of the model evaluation measures for which to calculate optimal thresholds. The default is using all measures available in 'modEvAmethods("threshMeasures")'.

optimize

character vector indicating the threshold optimization criteria to use; "each" calculates the optimal threshold for each model evaluation measure, while the remaining options optimize all measures according to the specified criterion. The default is using all criteria available in 'modEvAmethods("optiThresh")'.

simplif

logical, whether to calculate a faster simplified version. Used internally in other functions.

plot

logical, whether to plot the values of each evaluation measure at all thresholds.

sep.plots

logical. If TRUE, each plot is presented separately (you need to be recording R plot history to be able to browse through them all); if FALSE(the default), all plots are presented together in the same plotting window.

xlab

character vector indicating the label of the x axis.

…

additional arguments to pass to plot.

Value

This function returns a list with the following components:

all.thresholds

a data frame with the values of all analysed measures at all analysed thresholds.

optimals.each

if "each" is among the threshold criteria specified in 'optimize', optimals.each is output as a data frame with the value of each measure at its optimal threshold, as well as the type of optimal for that measure (which may be the maximum for measures of goodness such as "Sensitivity", or the minimum for measures of badness such as "Omission").

optimals.criteria

a data frame with the values of measure at the threshold that maximizes each of the criteria specified in 'optimize' (except for "each", see above).

References

Jimenez-Valverde A. & Lobo J.M. (2007) Threshold criteria for conversion of probability of species presence to either-or presence-absence. Acta Oecologica 31: 361-369.

Liu C., Berry P.M., Dawson T.P. & Pearson R.G. (2005) Selecting thresholds of occurrence in the prediction of species distributions. Ecography 28: 385-393.

Nenzen H.K. & Araujo M.B. (2011) Choice of threshold alters projections of species range shifts under climate change. Ecological Modelling 222: 3346-3354.

Examples

Run this code

# NOT RUN {
# load sample models:
data(rotif.mods)

# choose a particular model to play with:
mod <- rotif.mods$models[[1]]

# }
# NOT RUN {
optiThresh(model = mod)
# }
# NOT RUN {

# change some of the parameters:

optiThresh(model = mod, pch = 20, 
measures = c("CCR", "Sensitivity", "kappa", "TSS"), ylim = c(0, 1))


# you can also use optiThresh with vectors of observed and predicted
# values instead of with a model object:

# }
# NOT RUN {
optiThresh(obs = mod$y, pred = mod$fitted.values, pch = ".")
# }