Skip to contents

Linear mixed models

Source: R/stats.R
perform_lmer.Rd

Fits a linear mixed model separately for each feature. Returns all relevant statistics.

Usage

perform_lmer(
  object,
  formula_char,
  all_features = FALSE,
  ci_method = c("Wald", "profile", "boot"),
  test_random = FALSE,
  assay.type = NULL,
  ...
)

Arguments

object

a SummarizedExperiment or MetaboSet object

formula_char

character, the formula to be used in the linear model (see Details)

all_features

should all features be included in FDR correction?

ci_method

The method for calculating the confidence intervals as in confint

test_random

logical, whether tests for the significance of the random effects should be performed

assay.type

character, assay to be used in case of multiple assays

...

additional parameters passed to lmer

Value

A data frame with one row per feature, with all the relevant statistics of the linear mixed model as columns.

Details

The model is fit on combined_data(object). Thus, column names in pheno data can be specified. To make the formulas flexible, the word "Feature" must be used to signal the role of the features in the formula. "Feature" will be replaced by the actual Feature IDs during model fitting, see the example. With bootstrap ("boot") confidence intervals, the results are reproducible if RNGseed is set for the BiocParallel backend.

See also

lmer for model specification

Examples

data(example_set)
# A simple example without QC samples
# Features predicted by Group and Time as fixed effects with Subject ID as a 
# random effect
lmer_results <- perform_lmer(drop_qcs(example_set),
  formula_char = "Feature ~ Group + Time + (1 | Subject_ID)",
  ci_method = "Wald"
)
#> INFO [2025-06-23 22:37:27] Starting fitting linear mixed models.
#> INFO [2025-06-23 22:37:30] Linear mixed models fit.