poppr (version 2.9.6)

diversity_boot: Perform a bootstrap analysis on diversity statistics

Description

This function is intended to perform bootstrap statistics on a matrix of multilocus genotype counts in different populations. Results from this function should be interpreted carefully as the default statistics are known to have a downward bias. See the details for more information.

Usage

diversity_boot(
  tab,
  n,
  n.boot = 1L,
  n.rare = NULL,
  H = TRUE,
  G = TRUE,
  lambda = TRUE,
  E5 = TRUE,
  ...
)

Value

a list of objects of class "boot".

Arguments

tab

a table produced from the poppr function mlg.table(). MLGs in columns and populations in rows

n

an integer > 0 specifying the number of bootstrap replicates to perform (corresponds to R in the function boot::boot().

n.boot

an integer specifying the number of samples to be drawn in each bootstrap replicate. If n.boot < 2 (default), the number of samples drawn for each bootstrap replicate will be equal to the number of samples in the data set.

n.rare

a sample size at which all resamplings should be performed. This should be no larger than the smallest sample size. Defaults to NULL, indicating that each population will be sampled at its own size.

H

logical whether or not to calculate Shannon's index

G

logical whether or not to calculate Stoddart and Taylor's index (aka inverse Simpson's index).

lambda

logical whether or not to calculate Simpson's index

E5

logical whether or not to calculate Evenness

...

other parameters passed on to boot::boot() and diversity_stats().

Author

Zhian N. Kamvar

Details

Bootstrapping is performed in three ways:

• if n.rare is a number greater than zero, then bootstrapping is performed by randomly sampling without replacement n.rare samples from the data.

  \item if `n.boot` is greater than 1, bootstrapping is performed by
  sampling n.boot samples from a multinomial distribution weighted by the
  proportion of each MLG in the data.
  \item if `n.boot` is less than 2, bootstrapping is performed by 
  sampling N samples from a multinomial distribution weighted by the
  proportion of each MLG in the data.
  

Downward Bias

When sampling with replacement, the diversity statistics here present a downward bias partially due to the small number of samples in the data. The result is that the mean of the bootstrapped samples will often be much lower than the observed value. Alternatively, you can increase the sample size of the bootstrap by increasing the size of n.boot. Both of these methods should be taken with caution in interpretation. There are several R packages freely available that will calculate and perform bootstrap estimates of Shannon and Simpson diversity metrics (eg. entropart, entropy, simboot, and EntropyEstimation. These packages also offer unbiased estimators of Shannon and Simpson diversity. Please take care when attempting to interpret the results of this function.

See Also

diversity_stats() for basic statistic calculation, diversity_ci() for confidence intervals and plotting, and poppr(). For bootstrap sampling: stats::rmultinom() boot::boot()

Examples

library(poppr)
data(Pinf)
tab <- mlg.table(Pinf, plot = FALSE)
diversity_boot(tab, 10L)
if (FALSE) {
# This can be done in a parallel fashion (OSX uses "multicore", Windows uses "snow")
system.time(diversity_boot(tab, 10000L, parallel = "multicore", ncpus = 4L))
system.time(diversity_boot(tab, 10000L))
}