Skip to contents

Phylogenetic partial ROC (Peterson, Papes & Soberon 2008 analogue)

Source: R/phylo_partial_roc.R
phylo_partial_roc.Rd

Partial ROC test for presence-only host-suitability predictions, following Peterson, Papes & Soberon (2008, Ecological Modelling 213, 63-72). The ROC space is redefined for presence-only data: the y axis is the sensitivity \(a(t)\) (proportion of known hosts predicted at cutoff \(t\)) and the x axis is the proportion of ALL species in the phylogeny predicted as hosts, \(b(t)\) (the analogue of proportional predicted area). The curve is restricted to the low-omission region \(a(t) \ge 1 - E\) for a user-chosen omission tolerance error_rate \(E\), and performance is summarized as the ratio of the partial AUC of the observed curve to the partial AUC of the null (random-ranking) expectation, the diagonal \(a = b\). A ratio of 1 equals random performance; ratios above 1 indicate predictive signal. Significance is assessed by bootstrap resampling of the known hosts.

Usage

phylo_partial_roc(
  pred,
  is_host,
  error_rate = 0.05,
  n_boot = 500,
  boot_prop = 0.5,
  seed = NULL
)

Arguments

pred

Numeric vector of predicted suitability scores, one per species in the phylogeny.

is_host

Logical or 0/1 vector of the same length: TRUE/1 for known hosts.

error_rate

Allowed omission error \(E\) in \([0, 1)\). The partial AUC is computed over the region with sensitivity \(\ge 1 - E\). Default 0.05.

n_boot

Number of bootstrap replicates. Default 500.

boot_prop

Proportion of known hosts resampled (with replacement) in each bootstrap replicate, as in Peterson, Papes and Soberon (2008). Default 0.5.

seed

Integer or NULL. Random seed for the bootstrap. Default NULL.

Value

An object of class "phylopred_partial_roc", a list with:

auc_ratio

Observed partial AUC ratio (observed / null).

p_value

Bootstrap proportion of replicates with ratio \(\le 1\).

boot_ratios

Vector of bootstrap AUC ratios.

curve

data.frame with threshold, a (sensitivity) and b (predicted breadth).

error_rate, n_boot, boot_prop

The configuration used.

See also

host_threshold_metric

Examples

pairs <- prepare_pair_data(beetleTreeInteractions, phy_dist)
#> Warning: 22 interaction record(s) dropped: host not in 'phydist'.
fit <- stats::glm(suscept ~ phydist, data = pairs, family = stats::binomial())
one <- pairs[pairs$parasite == pairs$parasite[1], ]
one <- one[!duplicated(one$target), ]
pred <- stats::predict(fit, newdata = one, type = "response")
phylo_partial_roc(pred, one$suscept, n_boot = 50, seed = 42)
#> Phylopred phylogenetic partial ROC (E = 0.05)
#> Partial AUC ratio: 1.0284
#> Bootstrap p-value (ratio <= 1): 0.0000  [50 replicates]