occurrence.threshold: Test occurrence probability thresholds

In jeffreyevans/rfUtilities: Random Forests Model Selection and Performance Evaluation

Test occurrence probability thresholds

Description

A statistical sensitivity test for occurrence probability thresholds

Usage

occurrence.threshold(
  x,
  class,
  p = seq(0.1, 0.9, 0.05),
  type = c("delta.ss", "sum.ss", "kappa", "youden", "logloss")
)

Arguments

x	A classification randomForest model object
class	What class to test, quoted
p	Vector of probability thresholds
type	What statistic to use in evaluation ("delta.ss", "sum.ss", "kappa", "youden", "logloss")

Details

Available threshold evaluation statistics:

• kappa - [The Kappa statistic is maximized]

• sum.ss - [The sum of sensitivity and specificity is maximized]

• delta.ss - [The absolute value of the difference between sensitivity and specificity is minimized]

• youden - [Youden’s J index (sensitivity + specificity) - 1]

• logloss - [Optimizes on Log Loss]

Value

An "occurrence.threshold" class object containing a "thresholds" vector, object with evaluation statistic and probability thresholds as names, the probability threshold "prob.threshold" and the test statistic

Author(s)

Jeffrey S. Evans <jeffrey_evans<at>tnc.org>

References

Jimenez-Valverde, A., & J.M. Lobo (2007). Threshold criteria for conversion of probability of species presence to either-or presence-absence. Acta Oecologica 31(3):361-369

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

Youden, W.J. (1950). Index for rating diagnostic tests. Cancer. 3:2-35.

Examples

library(randomForest)
 data(imports85)
  imp85 <- imports85[,-2] 
  imp85 <- imp85[complete.cases(imp85), ]
  imp85[] <- lapply(imp85, function(x) if (is.factor(x)) x[, drop=TRUE] else x)
  y <- ifelse( imp85$numOfDoors != "four", "0", "1")   
  
( rf.mdl <- randomForest(y = as.factor(y), x = imp85[,-5]) )
  ( delta.ss <- occurrence.threshold(rf.mdl, class = "1", , type = "delta.ss") )
  ( sum.ss <- occurrence.threshold(rf.mdl, class = "1", type = "sum.ss") ) 
  ( kappa.ss <- occurrence.threshold(rf.mdl, class = "1", type = "kappa") )
  ( logloss.ss <- occurrence.threshold(rf.mdl, class = "1", type = "logloss") )
  ( youden.ss <- occurrence.threshold(rf.mdl, class = "1", type = "youden") )
 opar <- par  
   par(mfrow=c(2,2))
     plot(sum.ss)
     plot(delta.ss)
     plot(kappa.ss)   
     plot(logloss.ss)
 par <- opar
 
#### using ranger
library(ranger) 
( rf.mdl <- ranger(y = as.factor(y), x = imp85[,-5], probability = TRUE) )
 ( kappa.ss <- occurrence.threshold(rf.mdl, class = "1", type = "kappa") )
 ( youden.ss <- occurrence.threshold(rf.mdl, class = "1", type = "youden") )
  plot(kappa.ss) 

jeffreyevans/rfUtilities documentation built on Nov. 12, 2023, 6:52 p.m.