Representational connectivity model (ReNA-RC)

For each ROI or searchlight, computes representational connectivity between the ROI RDM and a seed RDM, optionally controlling for confound RDMs (block, lag, behavior, etc.).

Usage

repnet_model(
  dataset,
  design,
  repnet_des,
  distfun = cordist(method = "pearson"),
  simfun = c("pearson", "spearman"),
  ...
)

Arguments

dataset

An mvpa_dataset.

design

An mvpa_design.

repnet_des

Output of repnet_design.

distfun

Distance function for ROI RDM (e.g. cordist(method = "pearson")).

simfun

Character: similarity metric between ROI and seed RDMs ("pearson" or "spearman").

...

Extra fields stored on the model spec.

Value

A model spec of class "repnet_model" compatible with run_regional() and run_searchlight().

Details

Per-ROI metrics: n_items, n_pairs, conn_raw (the ROI-seed RDM similarity), and from the multiple regression of the ROI RDM on the seed and confound RDMs, beta_<term> (regression coefficients) and sp_<term> (semi-partial correlations). With confound RDMs, seed_conf_r2 is the R-squared of the seed RDM regressed on the confound RDMs: beta_seed and sp_seed describe only the 1 - seed_conf_r2 share of the seed that the confounds do not explain.

Examples

ds <- gen_sample_dataset(D = c(4, 4, 4), nobs = 16, blocks = 4, nlevels = 2)
# Seed RDM over trials (e.g. from a seed region or a behavioural model)
key_ids <- ds$design$train_design$.rownum
seed_rdm <- as.matrix(dist(key_ids))
rownames(seed_rdm) <- colnames(seed_rdm) <- as.character(key_ids)
rn_des <- repnet_design(ds$design, key_var = ~ .rownum, seed_rdm = seed_rdm)
model <- repnet_model(ds$dataset, ds$design, rn_des)