Struct GaugeComposedJacobian 

pub struct GaugeComposedJacobian { /* private fields */ }

A BlockEffectiveJacobian that composes an inner callback’s raw-width effective Jacobian with a fixed reduced→raw block transform T_b (p_raw × r_reduced), so the family sees the reduced coordinates by construction (#933).

The inner callback emits its row Jacobian in the raw coordinate system ((rows · k) × p_raw), the layout every BlockEffectiveJacobian impl produces — channel-major rows, raw columns. Post-multiplying each row by T_b rotates those raw columns into the reduced section: the effective reduced Jacobian is J_raw · T_b, with r_reduced columns. On the model η = J_raw · β_raw = (J_raw · T_b) · θ this is the exact reduced operator for the reduced coefficient θ, and the family lifts θ back to β_raw through the SAME T_b via the one Gauge.

This is the inversion #933 calls for: instead of forwarding a raw-width callback alongside a column-selection T_i (which leaves the family asserting raw column counts on a reduced spec and panicking), the callback is wrapped so its output already has the reduced width — the family captures the reduced design and its row-Hessian column-count assertions hold by construction. A column-selection T_b (zero/one entries) makes this exactly the audit’s drop; a general orthonormal T_b makes it any gauge section.

Implementations

impl GaugeComposedJacobian

pub fn new( inner: Arc<dyn BlockEffectiveJacobian>, t_block: Arc<Array2<f64>>, ) -> Self

Wrap inner so its effective Jacobian is post-multiplied by t_block (p_raw × r_reduced). t_block.nrows() must equal the inner callback’s raw column count.

Trait Implementations

impl BlockEffectiveJacobian for GaugeComposedJacobian

fn effective_jacobian_rows( &self, state: &FamilyLinearizationState<'_>, rows: Range<usize>, ) -> Result<Array2<f64>, String>

Stacked multi-output Jacobian for a contiguous observation row range.

fn n_outputs(&self) -> usize

Number of stacked output channels. 1 for most blocks.

fn eta_row_scaling_for_skewness(&self) -> Option<Arc<[f64]>>

Returns the per-row scaling vector when this callback is a simple diagonal-scaling block (RowScaledJacobian). Used by the identifiability audit’s skewness-aware bias correction (T25).

fn effective_jacobian_at( &self, state: &FamilyLinearizationState<'_>, ) -> Result<Array2<f64>, String>

Stacked multi-output Jacobian at the current β.