Skip to contents

boot_sample Generate n bootstrap sample of the monitoring sites to be used for each iteration.

Source: R/rbms_index_modelling.r
boot_sample.Rd

boot_sample Generate n bootstrap sample of the monitoring sites to be used for each iteration.

Usage

boot_sample(
  data,
  boot_n = 1000,
  s_sp = NULL,
  sindex_value = "SINDEX",
  strata = TRUE,
  check_link = TRUE,
  min_year = 3,
  max_trial = 5,
  habitat = NULL
)

Arguments

data

data.table or data.frame Data with all site id, with SPECIES, M_YEAR and sindex_value when s_sp is given.

boot_n

integer The number of bootstrap samples to be generated.

s_sp

string Species for which the bootstrap samples are drawn, default NULL draws among all sites without checks.

sindex_value

string Name of the site index variable used to define strata and observations, default is "SINDEX".

strata

logical Stratify the sampling by the occurrence of zeros at the sites and by group of linked years, default TRUE.

logical Draw again the samples where a group of linked years is split or lost, default TRUE.

min_year

integer Minimum number of years in a bootstrap sample, default 3.

max_trial

integer Maximum number of trials to draw a valid bootstrap sample, default 5.

habitat

string Name of a column of data with the habitat (or stratum) of each site, constant within site, default NULL. The habitat is crossed with the strata, so each sample keeps the number of sites of each habitat (design stratified by habitat, see collated_index_mixed).

Value

A list with site id and bootstrap indices for n bootstrap sample. When s_sp is given, the list also contains the species (s_sp), the number of groups of years of the original data (n_comp_0), the site indices per stratum (strata_ind, named group_zero pattern), the group of each year (year_group) and of each site (site_group), the number of trials per sample (n_trial) and the number of failed samples (n_failed).

Details

When s_sp is given, the samples are drawn for that species only, stratified by the occurrence of zeros at the sites (never zero, some zeros, always zero), and a sample is drawn again, up to max_trial trials, when it has fewer than min_year years or when its years are not linked through the sites with observations (see n_components). A sample still failing after max_trial trials is returned as a row of NA. When the years of the original data form several groups not linked through sites with observations (see collated_index), the strata are crossed with the groups (a site with observations belongs to one group, a site never observed to the group of most of its years), so each sample keeps the number of sites of each group, and each group must remain linked in the sample (not split or lost). With one group, this is the stratification by zeros only. For reproducible samples, call set.seed() before boot_sample(); when looping over species, setting the seed before each species makes its samples independent of the order or the subset of species processed.

See also

impute_count, collated_index

Author

Reto Schmucki - retoshm@ceh.ac.uk