taxatree_stats_p_adjust: Adjust p values in taxatree_stats dataframe

In david-barnett/microViz: Microbiome Data Analysis and Visualization

Adjust p values in taxatree_stats dataframe

Description

Apply a p value adjustment method from stats::p.adjust.methods, such as false-discovery rate adjustment with "BH", or more conservative family-wise error rate controlling methods such as "holm" or "bonferroni".

Usage

taxatree_stats_p_adjust(
  data,
  method,
  grouping = "rank",
  p = "p.value",
  new_var = NULL
)

Arguments

data	psExtra with taxatree_stats dataframe, or just the dataframe
method	any method from stats::p.adjust.methods
grouping	defines grouping of p-values into families for adjustment, see details.
p	name of variable containing p values for adjustment
new_var	name of new variable created for adjusted p values (automatically inferred by default)

Details

Define how to group the p values for adjustment with the grouping argument. The default is to adjust the p values in groups at each taxonomic rank, but you could also adjust per "taxon" or per "term". Or even group by a combination of rank and term with c("rank", "term"). You should specify the name of the new variable containing the adjusted p values in the new_var argument. If left as NULL the new variable name will be created by pasting together p.adj, the method, and the grouping variable(s) separated by ".".

Value

psExtra with dataframe of statistics, or just the data.frame

See Also

taxatree_models2stats

taxatree_models

stats::p.adjust

Examples

# This example is an abbreviated excerpt from article on taxon modelling on
# the microViz documentation website

library(corncob)
library(dplyr)
data("ibd", package = "microViz")

# We'll keep only the Ulcerative Colitis and Healthy Control samples, to
# simplify the analyses for this example. We'll also remove the Species
# rank information, as most OTUs in this dataset are not assigned to a
# species. We'll also use `tax_fix` to fill any gaps where the Genus is
# unknown, with the family name or whatever higher rank classification is
# known.

phylo <- ibd %>%
  ps_filter(DiseaseState %in% c("UC", "nonIBD")) %>%
  tax_mutate(Species = NULL) %>%
  tax_fix()

# Let's make some sample data variables that are easier to use and compare
# in the statistical modelling ahead. We will convert dichotomous
# categorical variables into similar binary variables (values: 1 for true,
# or 0 for false). We will also scale and center the numeric variable for
# age.

phylo <- phylo %>%
  ps_mutate(
    UC = ifelse(DiseaseState == "UC", yes = 1, no = 0),
    female = ifelse(gender == "female", yes = 1, no = 0),
    antibiotics = ifelse(abx == "abx", yes = 1, no = 0),
    steroids = ifelse(steroids == "steroids", yes = 1, no = 0),
    age_scaled = scale(age, center = TRUE, scale = TRUE)
  )

bb_models <- phylo %>%
  tax_fix() %>%
  tax_prepend_ranks() %>%
  tax_filter(min_prevalence = 0.3) %>%
  taxatree_models(
    type = corncob::bbdml,
    ranks = c("Phylum", "Class", "Order"),
    variables = c("UC", "female", "antibiotics", "steroids", "age_scaled")
  )

bb_stats <- bb_models %>%
  taxatree_models2stats(param = "mu") %>%
  taxatree_stats_p_adjust(method = "BH", grouping = "rank")

bb_stats

bb_stats %>% taxatree_stats_get()

# you can also directly modify the dataframe,
# and choose a different variable name
bb_stats %>%
  taxatree_stats_get() %>%
  taxatree_stats_p_adjust(
    method = "holm", grouping = "taxon", new_var = "p_adj_holm"
  )

# see all available adjustment methods
stats::p.adjust.methods

david-barnett/microViz documentation built on April 17, 2025, 4:25 a.m.