Skip to contents

Leaves n_test simulations out of the reference table as pseudo-observations with known parameters, fits abc_fit() on the rest and compares the estimates with the truth. A parameter that is not recovered from simulated data should not be inferred from real data (Pontarp et al. 2019).

Usage

abc_recovery(
  table,
  parameters = NULL,
  statistics = NULL,
  n_test = 200,
  level = 0.9,
  ntree = 500,
  ncores = 1,
  seed = 1
)

Arguments

table

An mc_abc_table from abc_table(), or a data frame with the parameters and the statistics.

parameters

Names of the parameters to estimate. By default those of table.

statistics

Names of the statistics to use. By default all the others.

n_test

Number of pseudo-observations.

level

Probability of the credible interval.

ntree

Number of trees per forest.

ncores

Number of threads for the forests.

seed

Integer for the choice of the pseudo-observations.

Value

A data frame with one row per parameter: n, r2 (1 minus the squared error over the variance of the truth), bias and width (mean width of the interval) as fractions of the prior range, and coverage, the fraction of intervals that contain the truth, with its standard error. The attribute estimates holds every estimate. A coverage near level says the intervals are calibrated; R2 says how much the data narrow the prior.

References

Pontarp, M., Brännström, Å., & Petchey, O. L. (2019). Inferring community assembly processes from macroscopic patterns using dynamic eco-evolutionary models and Approximate Bayesian Computation (ABC). Methods in Ecology and Evolution, 10, 450–460. doi:10.1111/2041-210X.13129

Examples

set.seed(1)
table <- data.frame(a = runif(400), b = runif(400))
table$s1 <- table$a + rnorm(400, 0, 0.05)
table$s2 <- rnorm(400)
abc_recovery(table, parameters = c("a", "b"), n_test = 50, ntree = 100)
#>   parameter  n         r2        bias coverage coverage_se    width
#> 1         a 50  0.9523572  0.01580127     0.86  0.04907138 0.223698
#> 2         b 50 -0.1949136 -0.01829474     0.70  0.06480741 0.798200