Module bms

Re-exports§

BernoulliMarginalSlopeAloRowGeometry: Negative-log-likelihood derivatives in the affine fitted coordinates [marginal eta, slope] for one saved Bernoulli marginal-slope row.
BernoulliMarginalSlopeAloRowInput: Complete saved-row state for exact rigid Bernoulli marginal-slope ALO replay.
BernoulliMarginalSlopeFitResult
BernoulliMarginalSlopeSavedAloReplay: Row-aligned replay of the fitted Bernoulli marginal-slope likelihood.
BernoulliMarginalSlopeSavedAloRowGeometry: Exact saved-row geometry in the full local primary frame [marginal eta, slope, score-warp coefficients..., link-deviation coefficients...].
BernoulliMarginalSlopeTermSpec
BmsLogslopeJacobian: β-dependent Jacobian for the BMS logslope block.
BmsMarginalJacobian: β-dependent Jacobian for the BMS marginal block.
CrossBlockIdentifiabilityWarning: Structured warning surfaced by the BMS family when a candidate flex block is fully aliased by its anchor union and gets dropped.
DeviationBlockConfig
EmpiricalZGrid
LatentZConditionalCalibration: Conditional location-scale calibration of the latent score (#905).
LatentZNormalization
LatentZPolicy
LatentZRankIntCalibration: Weighted mid-distribution rank inverse-normal transform for the latent score.
MarginalSlopeCovariance: Immutable, validated covariance geometry for the physical log-slope vector.

LatentMeasureCalibration: Optional calibration applied to the latent score before the BMS kernel runs. When RankInverseNormal, both the training and predict paths route the input z through [LatentZRankIntCalibration::apply_*] before the standard-normal closed-form kernel is invoked.
LatentMeasureKind
LatentMeasureSpec
LatentZCheckMode
LatentZNormalizationMode
MarginalSlopeCovarianceShape

bernoulli_marginal_slope_alo_row_geometry: Replay the exact rigid standard-normal row program used by fitting.
fit_bernoulli_marginal_slope_terms
marginal_slope_covariance_from_scores
marginal_slope_preserving_scale
marginal_slope_probit_eta
padded_deviation_seed