Variance decomposition by linear mixed models (LMM), to estimate
the contribution of Group and Time to the variance of each feature.
Modified from PALMO::lmeVariance() function.
Group and Time are always included as random effects. The formula is extended automatically based on colData:
Subject column present: adds
(1|Subject)to model between-subject differencesinteraction = TRUE: adds(1|Group:Time)(or(1|Subject:Time)when Subject present) to capture interaction variance. Requires >= 2 replicates per combination. Default: FALSE.
Output always includes: Group, Time, Residual. Subject is included when present. All values are percentages of total variance.
Allow using original data or data normalised to time point 0.
Usage
decomp_variance(
se_obj,
features = NULL,
fixed_effect_var = NULL,
interaction = FALSE,
assay,
core = 1
)Arguments
- se_obj
A SummarizedExperiment object created by
create_input()- features
A vector of features to include. If NULL, all features.
- fixed_effect_var
Column names to include as fixed effects (regressed out, not appearing as variance components). Default is
NULL(no fixed effects). Set to"Batch"to regress out batch effects, or provide other columns (e.g.,c("Sex", "Age")).- interaction
Logical. If TRUE, adds
(1|Group:Time)(or(1|Subject:Time)when Subject present) to the model to capture interaction variance. Default: FALSE.- assay
Assay to use:
"orig"for original data,"norm"for normalised-to-start data. Numeric indices (1, 2) are also accepted. No default, must be specified explicitly.- core
Number of cores for parallel processing (default: 1)
Examples
data(example_obj)
example_obj <- normalise_to_start(example_obj)
#> Normalising to group baseline at each feature's first non-NA time point.
var_decomp <- decomp_variance(example_obj, assay = 1)
#> LMM: exp ~ (1|Group) + (1|Time) | Output: Group, Time, Residual