test.dispRity {dispRity}

Testing disparity hypotheses

Description

Applying statistical tests to dispRity objects

Usage

test.dispRity(
  data,
  test,
  comparisons = "pairwise",
  rarefaction = NULL,
  correction = "none",
  concatenate = TRUE,
  conc.quantiles = c(mean, c(95, 50)),
  details = FALSE,
  ...
)

Arguments

data

A dispRity object.

test

A test function to apply to the data.

comparisons

If data contains more than two subsets, the type of comparisons to apply: either "pairwise" (default), "referential", "sequential", "all" or a list of pairs of subset names/number to compare (see details).

rarefaction

A numeric value indicating whether to use a specific rarefaction level (default = NULL).

correction

Which p-value correction to apply to htest category test (see p.adjust; default = "none").

concatenate

Logical, whether to concatenate bootstrapped disparity values (TRUE; default) or to apply the test to each bootstrapped value individually (FALSE).

conc.quantiles

If concatenate = TRUE, must be a central tendency function and a vector of quantiles (default = c(mean, c(95, 50))).

details

Whether to output the details of each test (non-formatted; default = FALSE).

...

Additional options to pass to the test function.

Details

The comparison argument can be:

• "pairwise": pairwise comparisons of all the subsets (default).

• "referential": compares the first subset to all the others.

• "sequential": compares each subset sequentially (e.g. first against second, second against third, etc.).

• "all": compares all the subsets simultaneously to the data (i.e. bootstrapped disparity ~ subsets names). This argument is used for lm or glm type tests.

• A list of pairs of number of subsets to compare. Each element of the list must contain two elements (e.g. list(c("a","b"), ("b", "a")) to compare "a" to "b" and then "b" to "a").

• If the called test is null.test, the comparison argument is ignored.

IMPORTANT: if you are performing multiple comparisons (e.g. when using "pairwise", "referential" or "sequential"), don't forget about the Type I error rate inflation. You might want to use a p-value correction (see p.adjust).

Author(s)

Thomas Guillerme

See Also

dispRity, null.test, bhatt.coeff, pair.plot, adonis.dispRity, randtest.dispRity, test.dispRity

Examples

## Load the Beck & Lee 2014 data
data(BeckLee_mat50)
data(BeckLee_tree)

## Calculating the disparity from customised subsets
## Generating the subsets
groups <- crown.stem(BeckLee_tree, inc.nodes = FALSE)
customised_subsets <- custom.subsets(BeckLee_mat50, groups)
## Bootstrapping the data
bootstrapped_data <- boot.matrix(customised_subsets, bootstraps = 100)
## Calculating the sum of variances
sum_of_variances <- dispRity(bootstrapped_data, metric = c(sum, variances))

## Measuring the subset overlap
test.dispRity(sum_of_variances, bhatt.coeff, "pairwise")

## Measuring differences from a reference subset
test.dispRity(sum_of_variances, wilcox.test, "referential")

## Measuring disparity as a distribution
disparity_var <- dispRity(bootstrapped_data, metric = variances)
## Differences between the concatenated bootstrapped values of the subsets
test.dispRity(disparity_var, test = t.test, comparisons = "pairwise",
     concatenate = TRUE, correction = "bonferroni")
## Differences between the subsets bootstrapped
test.dispRity(disparity_var, test = t.test, comparisons = "pairwise",
     concatenate = FALSE, correction = "bonferroni",
     conc.quantiles = c(mean, c(95, 5)))