occurrence.threshold: Test occurrence probability thresholds

In rfUtilities: Random Forests Model Selection and Performance Evaluation

Description

A statistical sensitivity test for occurrence probability thresholds

Usage

occurrence.threshold(x, xdata, class, p = seq(0.1, 0.7, 0.02),
  type = "delta.ss")

Arguments

x	A classification randomForest model object
xdata	Independent data used to build model
class	What class to test
p	Vector of probability thresholds
type	What statistic to use in evaluation ("delta.ss", "sum.ss", "kappa")

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

Value

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

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.

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.t <- occurrence.threshold(rf.mdl, imp85[,-5], class = "1") )
   ( sum.ss.t <- occurrence.threshold(rf.mdl, imp85[,-5], class = "1", 
                                      type = "sum.ss") ) 
   ( kappa.ss.t <- occurrence.threshold(rf.mdl, imp85[,-5], class = "1",
                                      type = "kappa") )
  
par(mfrow=c(2,2))
  plot(sum.ss.t)
  plot(delta.ss.t)
  plot(kappa.ss.t)   
  

Example output

rfUtilities 2.1-1
Type rfu.news() to see new features/changes/bug fixes.
randomForest 4.6-12
Type rfNews() to see new features/changes/bug fixes.

Call:
 randomForest(x = imp85[, -5], y = as.factor(y)) 
               Type of random forest: classification
                     Number of trees: 500
No. of variables tried at each split: 4

        OOB estimate of  error rate: 10.88%
Confusion matrix:
   0   1 class.error
0 67  14   0.1728395
1  7 105   0.0625000
Evaluation statistic: delta.ss 

Moments for thresholds: 

   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
0.00000 0.01786 0.04464 0.08415 0.12346 0.38272 

Probability threshold with minimum delta sensitivity/specificity =  0.46 
Evaluation statistic: sum.ss 

Moments for thresholds: 

   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
  1.617   1.877   1.955   1.916   1.982   2.000 

Probability threshold with maximum cummlative sensitivity/specificity =  0.46 
Evaluation statistic: kappa 

Moments for thresholds: 

   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
 0.6518  0.8918  0.9572  0.9224  0.9788  1.0000 

Probability threshold with maximum kappa =  0.46 

rfUtilities documentation built on Oct. 3, 2019, 9:04 a.m.