Skip to contents

Temporal nuisance RDM at condition level (for MS-ReVE)

temporal_nuisance_for_msreve.Rd

Creates a condition-level temporal nuisance RDM for use with MS-ReVE/contrast_rsa_model. This function reduces trial-level temporal relationships to condition-level relationships.

Usage

temporal_nuisance_for_msreve(
  mvpa_design,
  time_idx,
  reduce = c("min", "mean", "median", "nn_min"),
  kernel = c("adjacent", "boxcar", "linear", "poly", "exp", "gauss"),
  within_blocks_only = TRUE,
  ...
)

Arguments

mvpa_design

an mvpa_design object containing Y (condition labels) and block_var

time_idx

numeric/integer vector of temporal indices, length = nrow(mvpa_design$train_design)

reduce

character string specifying reduction method:

  • "min"Minimum lag between any pair of trials from two conditions

  • "mean"Average lag between all pairs of trials

  • "median"Median lag between all pairs of trials

  • "nn_min"Minimum nearest-neighbor distance

kernel

character string specifying kernel type (see temporal_rdm)

within_blocks_only

logical; if TRUE, ignore pairs spanning different blocks (default TRUE)

...

additional kernel parameters passed to kernel functions (width, power, lambda, sigma)

Value

K x K symmetric matrix (0 diagonal) aligned to levels(mvpa_design$Y)

Details

This function is designed for MS-ReVE analyses where temporal confounds need to be modeled at the condition level rather than trial level. It computes aggregate temporal relationships between conditions based on the temporal structure of individual trials.

Examples

if (FALSE) { # \dontrun{
# Create temporal nuisance for MS-ReVE
temp_K <- temporal_nuisance_for_msreve(
  mvpa_design = mvpa_des,
  time_idx = seq_len(nrow(mvpa_des$train_design)),
  reduce = "min",
  kernel = "exp", 
  lambda = 3,
  within_blocks_only = TRUE
)

# Use in msreve_design
msreve_des <- msreve_design(
  mvpa_design = mvpa_des,
  contrast_matrix = C_mat,
  nuisance_rdms = list(temp_decay = temp_K)
)
} # }