Automated model selection for Hill diversity on each phyloseq in a list_phyloseq

Performs automated model selection and multimodel inference using MiscMetabar::glmutli_pq() on each phyloseq object in a list_phyloseq. Returns a summary table with the results from all phyloseq objects.

Usage

glmulti_lpq(
  x,
  formula,
  fitfunction = "lm",
  q = c(0, 1, 2),
  aic_step = 2,
  confsetsize = 100,
  plotty = FALSE,
  level = 1,
  method = "h",
  crit = "aicc",
  verbose = TRUE,
  ...
)

Arguments

x: (required) A list_phyloseq object.
formula: (character, required) A model formula for glmulti. Variables must be present in the sample_data slot of all phyloseq objects. Hill numbers (Hill_0, Hill_1, Hill_2) and Abundance are automatically available.
fitfunction: (character, default "lm") The model fitting function to use. Options include "lm" for linear models or "glm" for generalized linear models.
q: (numeric vector, default c(0, 1, 2)) The q values for Hill number computation. Defaults to Hill numbers 0 (richness), 1 (Shannon exponential), and 2 (inverse Simpson).
aic_step: (numeric, default 2) The AIC score threshold for model selection. Models within this threshold from the best model are included.
confsetsize: (integer, default 100) The number of models to return in the confidence set.
plotty: (logical, default FALSE) If TRUE, display IC profile during glmulti search.
level: (integer, default 1) Model complexity level. 1 for main effects only, 2 for pairwise interactions.
method: (character, default "h") The search method for glmulti. Options: "h" (exhaustive), "g" (genetic algorithm), "l" (branch-and-bound), "d" (summary only).
crit: (character, default "aicc") Information criterion for model selection. Options include "aic", "aicc" (small-sample corrected AIC), "bic".
verbose: (logical, default TRUE) If TRUE, print progress messages.
...: Additional arguments passed to MiscMetabar::glmutli_pq().

Value

A tibble with the combined results from all phyloseq objects, containing the following columns:

name: Name of the phyloseq object
variable: The variable name from the model
estimates: The model coefficient estimate
unconditional_interval: Confidence interval from model averaging
nb_model: Number of models containing this variable
importance: Relative importance of the variable (sum of Akaike weights)
alpha: Significance level

Details

This function requires that the list_phyloseq type is NOT SEPARATE_ANALYSIS, as the formula must contain variables that are common across all phyloseq objects.

The function wraps MiscMetabar::glmutli_pq(), which itself wraps the glmulti package for automated model selection. For each phyloseq object, Hill diversity indices are computed and used as response variables in the model selection process.

Examples

if (FALSE) { # \dontrun{
lpq <- list_phyloseq(
  list(
    fungi = data_fungi,
    fungi_clust = postcluster_pq(data_fungi)
  ),
  same_bioinfo_pipeline = FALSE
)

results <- glmulti_lpq(lpq, formula = "Hill_0 ~ Height + Time")
results

# With interactions
results_int <- glmulti_lpq(lpq, formula = "Hill_1 ~ Height * Time", level = 2)
} # }