Performs block bootstrap resampling for multiresponse linear models, particularly useful for fMRI time series data where temporal dependencies exist. The function implements a block bootstrap approach to maintain the temporal correlation structure within the data.
Usage
multiresponse_bootstrap_lm(
form,
data_env,
conlist,
vnames,
fcon,
modmat = NULL,
block_size = 30,
boot_rows = FALSE,
nboot = 100,
event_indices
)Arguments
- form
Formula for the linear model. Required if modmat is NULL.
- data_env
Environment containing the data for the linear model.
- conlist
List of contrasts to be computed for each bootstrap sample.
- vnames
Vector of variable names.
- fcon
Contrasts for fixed effects.
- modmat
Optional pre-computed model matrix. If provided, form is ignored.
- block_size
Size of the blocks for the bootstrap (default: 30). Should be large enough to capture temporal dependencies but small enough to allow sufficient randomization.
- boot_rows
Logical flag indicating whether to bootstrap rows (default: FALSE).
- nboot
Number of bootstrap iterations (default: 100).
- event_indices
Indices of events for computing beta covariances.
Value
A list containing:
original: The fitted original model with contrasts
con_cov: Covariance matrices for contrasts (if contrasts provided)
beta_cov: Covariance matrices for beta estimates
nboot: Number of bootstrap iterations performed
bootstrap: Logical indicating this is a bootstrap result
Details
The function performs the following steps:
Fits the original linear model
Implements block bootstrap resampling of residuals
Reconstructs response variables using fitted values and resampled residuals
Computes contrasts for each bootstrap sample
The block bootstrap approach helps preserve temporal dependencies in the data by resampling blocks of consecutive observations rather than individual observations.