Run pairwise IMD as a standalone post-fit analysis. By default, this uses
the selected features from the variable-selection / robust-clustering stage
when available, rather than all variables.
Usage
pairwise_imd(
x,
feature_source = c("selected", "weights_nonzero", "all"),
normalized = FALSE
)
Arguments
- x
An mrf3_fit or mrf3 object containing net, connection, and
IMD weights.
- feature_source
Which feature set to use:
"selected" (default), "weights_nonzero", or "all".
- normalized
Logical; whether to return degree-normalized adjacency.
Value
A list with adj_var_mat, adj_dat_mat, var_use, and
feature_source.
Details
Pairwise IMD is only defined for two-block connections; fits with
single-block (self) connections are rejected with an error. Feature names
must also be unique across omics blocks.