COmpositional Nonlinear-debiased Inference, Fay–Herriot with Ellipsoidal conformal Regions
Robust. Compositional. Confident.

Design-aware small-area estimation of forest structure as distributions, not just totals. The first release estimates the diameter distribution — stem density split across DBH classes — for small forest areas where the field sample is too thin for a reliable direct estimate, and it comes with an honest, valid statement of uncertainty.

Under the hood it is a compositional area-level Fay–Herriot estimator with a cross-fitted, one-step-debiased machine-learned mean and design-aware conformal prediction sets on the simplex. It reduces exactly to classical Fay–Herriot when the mean is linear.

pip install conifer-sae

Quickstart

import numpy as np, conifer

# counts: m areas x K DBH classes ; area_eff: effective sampled area ; X: covariates
est = conifer.DiameterDistribution(seed=0).fit(counts, area_eff, X)
s_hat = est.s_hat_                                   # stem density by DBH class

# design-aware conformal prediction set (valid joint coverage on the simplex)
est.conformalize(s_truth_cal, cal_idx, joint=True, alpha=0.10)
lo, hi = est.predict_interval(joint=True)

# coherence with a design-based benchmark (e.g. FIA class totals)
est.benchmark(totals, var_totals=var_totals)

Run the worked example:

python examples/quickstart.py

The names

CONIFER is a package; you build estimators on a shared engine.

Name	What it is
conifer.DiameterDistribution	the estimator you fit (formerly StemDensityClassSAE)
conifer.CompositionalFH	the reusable engine underneath it — debiased-ML mean + FH + conformal simplex sets
conifer.SpeciesComposition	planned v0.3 sibling: species shares on the same engine

A new region is a run, not a name — DiameterDistribution(spatial=True, regen_aware=True) toggles capabilities; you don't fork the package per state.

Why it exists (what nothing else does)

Small-area estimation and conformal prediction are both mature — but not together, and not on the simplex, and not for forestry:

• emdi / sae (R) give area-level FH with parametric MSE, but no compositional target and no distribution-free prediction sets.
• MAPIE / crepes (Python) give conformal prediction, but no small-area borrowing and no simplex geometry.
• rFIA / FIESTA give design-based forest estimates, but do not model or borrow strength.

CONIFER sits in that gap: compositional SAE + design-aware conformal sets on the simplex, forestry-native — and it consumes design-based FIA estimates as its benchmark rather than competing with them.

Command line

Point it at three aligned CSV matrices:

conifer fit --counts counts.csv --area area.csv --aux aux.csv --out s_hat.csv

Validation

The estimator was developed and stress-tested in the St. Joe (Idaho) study: on the merchantable diameter distribution it significantly beats the direct estimator and a broad competitor slate (kNN, Weibull, MERF, SAEforest, BART-FH, Dirichlet-multinomial, KBAABB, a multivariate FH), is coherent with FIA (design-consistent), and its conformal set is the only uncertainty statement with tested valid joint coverage (0.944). With the zero-robust log-ratio default the joint set is also efficient (~80% tighter at unchanged coverage). It reduces to classical Fay–Herriot (verified ratio ≈ 1.0002).

Roadmap

• v0.1 — DiameterDistribution: debiased-ML compositional FH + conformal simplex sets + benchmarking (this release)
• v0.2 — capability flags: spatial=True (spatial random effect), regen_aware=True (understory sub-model)
• v0.3 — SpeciesComposition (species shares) on the same engine
• v0.4 — non-compositional metrics (volume, basal area, biomass) via a MultivariateFH sibling core
• v1.0 — R front door via reticulate (one engine, library(conifer) for the FIA/forestry audience)

Citing

If you use CONIFER, please cite it — see CITATION.cff, or use GitHub's "Cite this repository" button. Method framing: a debiased-ML compositional Fay–Herriot with design-aware simplex conformal prediction sets for forest diameter distributions.

License

MIT — see LICENSE. Developed at the University of Idaho (IFC-UIDAHO).

---

The engine in conifer/_engine/ is a vendored copy of the validated PSAE research code; fixes flow from the research source of record (see conifer/_engine/__provenance__.txt). Brand assets (contour style) are in assets/.