Build subject-level random-effects design matrices from a parsed RE spec

Consumes the canonical block representation produced by .normalize_re_input() (Phase 1) and emits the per-observation design matrices Z_q0 and Z_alpha that the Phase-2 generalized src/MixedDemand.h template will consume.

Usage

.tmb_build_z_matrices(re_parsed, data, id_var = "id")

Arguments

re_parsed

Output of .normalize_re_input().

data

Long-format data frame the model is fit on.

id_var

Subject id column (currently unused — reserved for future per-subject reductions).

Value

A list with components:

Z_q0

Numeric matrix, n_obs rows by re_dim_q0 columns.

Z_alpha

Numeric matrix, n_obs rows by re_dim_alpha columns.

re_dim_q0

Integer; total Q0 RE columns.

re_dim_alpha

Integer; total alpha RE columns.

Details

For each block, the helper extracts the model.matrix() columns corresponding to that block's terms_q0 / terms_alpha from data and stacks them column-wise. Order: block 1 q0 cols, block 2 q0 cols, ..., then block 1 alpha cols, etc.

Intercept-only blocks (terms_* is "(Intercept)") produce a column of 1s; factor-expanded blocks reproduce the model.matrix expansion.