Measures how far the ratio of predicted to expected presences rises with
suitability. A well calibrated model gives a value near 1, a random one near 0.
Usage
calculate_cbi(pos_scores, all_scores, window_width = 0.1, n_bins = 100)
Arguments
- pos_scores
Numeric suitability scores at presence points. NA is dropped.
- all_scores
Numeric suitability scores at all points, presence and
background together. NA is dropped.
- window_width
Width of the moving window, as a proportion of the score
range.
- n_bins
Number of window positions to evaluate.
Value
A single number in [-1, 1], the Spearman correlation between window
position and the predicted-to-expected ratio. Returns 0 when there are no
presence scores, when all scores are equal, or when the correlation is
undefined.
Examples
set.seed(1)
background <- runif(500)
presences <- rbeta(50, 4, 2) # presences sit at higher scores
calculate_cbi(presences, c(presences, background))
#> [1] 0.5549083