Skip to contents

Evaluate an SDM

Source: R/evaluate_sdm.R
evaluate_sdm.Rd

Returns various evaluation metrics from predicts::pa_evaluate() and flexsdm::sdm_eval().

Usage

evaluate_sdm(m, p_test, b_test, do_gc = FALSE, ...)

Arguments

m

SDM result within tune_sdm()

p_test

Presence test data generated within tune_sdm()

b_test

Background test data generated within tune_sdm()

do_gc

Logical. Run base::rm(list = ls) and base::gc() at end of function? Useful when running SDMs for many, many taxa, especially if done in parallel. Note, actually usees rm(list = ls(pattern = "^[^e$]")).

...

Passed to both terra::predict() and predicts::pa_evaluate()

Value

paModelEvaluation (see predicts::pa_evaluate()) with extra metrics from flexsdm::sdm_eval(): AUC as auc_po_flexsdm; BOYCE as CBI; CBI_rescale (CBI is -1 to 1, CBI_rescale is 0 to 1); and IMAE.

Examples


  out_dir <- file.path(system.file(package = "envSDM"), "examples")

  source(fs::path(out_dir, "tune_sdm_ex.R")) # make sure following prep file exists
#> Warning: Missing `trust` will be set to FALSE by default for RDS in 2.0.0.
#> Warning: Missing `trust` will be set to FALSE by default for RDS in 2.0.0.
#> Warning: Missing `trust` will be set to FALSE by default for RDS in 2.0.0.
#> Warning: Missing `trust` will be set to FALSE by default for RDS in 2.0.0.
#> Warning: There were 2 warnings in `dplyr::mutate()`.
#> The first warning was:
#> ℹ In argument: `tune = purrr::map(tune, rio::import)`.
#> Caused by warning:
#> ! Missing `trust` will be set to FALSE by default for RDS in 2.0.0.
#> ℹ Run dplyr::last_dplyr_warnings() to see the 1 remaining warning.
#> Warning: There were 2 warnings in `dplyr::mutate()`.
#> The first warning was:
#> ℹ In argument: `tune = purrr::map(tune, rio::import)`.
#> Caused by warning:
#> ! Missing `trust` will be set to FALSE by default for RDS in 2.0.0.
#> ℹ Run dplyr::last_dplyr_warnings() to see the 1 remaining warning.

  prep <- rio::import(fs::path(out_dir, "acaule", "prep.rds"))
#> Warning: Missing `trust` will be set to FALSE by default for RDS in 2.0.0.

  model <- tune_sdm(prep = prep
                    , out_dir = FALSE
                    , return_val = "object"
                    , algo = "rf"
                    , trees = 500
                    , mtry = 2
                    , nodesize = 1
                    , keep_model = TRUE
                    )
#> tuning acaule
#> out_dir is H:/temp/nige\Rtmpk9T0CA\file44e41a463367
#> rf tune
#> Joining with `by = join_by(algo)`
#> Joining with `by = join_by(metric)`
#> Joining with `by = join_by(algo, tune_args, trees, nodesize)`

  presences <- prep$blocks[prep$blocks$pa == 1 & prep$blocks$block != 1, ]
  background <- prep$blocks[prep$blocks$pa == 0 & prep$blocks$block != 1, ]

  evaluate_sdm(model$tune_rf$m[[1]]
               , p_test = presences
               , b_test = background
               )
#> @stats
#>    np   na prevalence auc   cor pcor   ODP auc_po auc_po_flexsdm   CBI
#> 1 458 4521      0.092   1 0.937    0 0.908      1              1 0.904
#>   CBI_rescale  IMAE
#> 1       0.952 0.972
#> 
#> @thresholds
#>   max_kappa max_spec_sens no_omission equal_prevalence equal_sens_spec
#> 1      0.93         0.618       0.618            0.092            0.69
#> 
#> @tr_stats
#>     treshold kappa  CCR  TPR  TNR  FPR  FNR  PPP  NPP  MCR  OR
#> 1          0     0 0.09    1    0    1    0 0.09  NaN 0.91 NaN
#> 2          0  0.15 0.54    1 0.49 0.51    0 0.17    1 0.46 Inf
#> 3          0   0.2 0.62    1 0.58 0.42    0 0.19    1 0.38 Inf
#> 4        ...   ...  ...  ...  ...  ...  ...  ...  ...  ... ...
#> 143        1  0.87 0.98 0.79    1    0 0.21    1 0.98 0.02 Inf
#> 144        1  0.87 0.98 0.79    1    0 0.21    1 0.98 0.02 Inf
#> 145        1     0 0.91    0    1    0    1  NaN 0.91 0.09 NaN