Compute statistical significance of alpha diversity between groups

Runs global (Kruskal-Wallis or one-way ANOVA) and pairwise (Wilcoxon or Tukey HSD) hypothesis tests on each (rank, metric) combination present in the output of compute_alpha(). Pairwise p-values are adjusted with p.adjust() and Cohen's d effect sizes are appended.

Usage

compute_alpha_significance(
  x,
  group_col = NULL,
  metric = NULL,
  global_test = c("kruskal", "anova"),
  pairwise_test = c("wilcoxon", "tukey"),
  p_adjust_method = "BH"
)

Arguments

x

A "phip_alpha_diversity" list (output of compute_alpha()) or a single alpha-diversity data frame.

group_col

Character scalar; name of the grouping column. Inferred from attr(x, "group_cols") when NULL (default).

metric

Character vector; subset of metrics to test. NULL (default) uses all numeric metric columns present in the data.

global_test

One of "kruskal" (Kruskal-Wallis, default) or "anova" (one-way ANOVA).

pairwise_test

One of "wilcoxon" (Wilcoxon rank-sum, default) or "tukey" (Tukey HSD from aov()).

p_adjust_method

Method passed to p.adjust(). Default "BH".

Value

A named list of class "phip_alpha_significance" with two tibbles:

$global

One row per (rank, metric): rank, metric, statistic, p_value, test.

$pairwise

One row per (rank, metric, pair): rank, metric, group1, group2, p_raw, p_adj, cohens_d, stars, test.

Attributes: group_col, global_test, pairwise_test, p_adjust_method, metrics, ranks.

Examples

if (FALSE) { # \dontrun{
alpha <- compute_alpha(phip_obj, group_cols = "group")
sig   <- compute_alpha_significance(alpha)
sig$global
sig$pairwise
} # }