Methodology

What this is not

This is a modeled estimate, not an appraisal or a quote from any specific seller. It combines government data (energy prices, climate, heating-fuel prevalence) with published state surveys and carries a wide, honestly-stated uncertainty range — actual local prices vary by species, seasoning, delivery distance, and season. It is not a substitute for getting an actual quote from a local firewood seller.

Where a county's estimate is influenced by user-submitted data, remember that submissions are self-reported and unverified — Wood Heat Index doesn't independently confirm any individual price.

The model, in plain language

Every county starts from a structural prior: a formula relating firewood price to heating degree days (colder climates, more demand) and the local share of homes that heat with wood (more wood-heat demand, more of a local market). Where a real price observation exists for that county's state, the estimate shifts toward it — the more observations and the more precisely they're measured, the more weight they get, following a standard closed-form Bayesian shrinkage rule (no simulation, no per-county hand-tuning).

0 counties are observed (a real, county-resolved price observation dominates the estimate), 26 are state-anchored (influenced by a real observation, but only at state resolution), and 3118 — the large majority — are structural-only: theory alone, no observation anywhere informs the number. That imbalance is the honest state of available data, not a bug to be tuned away.

Evidence tiers

Published prices are also rounded to a precision the underlying evidence actually justifies — wide, structural-only estimates round to the nearest $25; well-supported ones to the nearest $5. An interval whose top is more than double its bottom is flagged "wide uncertainty" rather than shown with normal visual weight.

User submissions

Every county page has a form for sharing a real price you paid. A human reviews every submission before it can affect any published estimate — there is no automatic acceptance path.

Accepted submissions for one county are pooled, not averaged in the naive sense: the estimate uses the median of that county's accepted prices, so a single mistaken or unusual submission can't single-handedly move the number. The pooled evidence also counts for less than an equivalent state government survey — submissions from one county tend to share a season and self-select (someone surprised by a high quote is more likely to submit than someone who paid a typical price), so five independent submissions for a county are treated as roughly equivalent to one state-level survey observation, not five.

A county needs at least 3 independent submissions before user-submitted evidence alone can move it into the observed tier — one or two submissions still count (they can move a county from structural-only to state-anchored), but aren't enough on their own to claim the estimate is well-supported. This threshold, like the model's other declared coefficients below, is a stated judgment call, not something derived from data — there isn't yet enough submission history to derive it from.

Model coefficients

fitted means solved from a real anchor (with a stated degrees of freedom); every other basis is a stated, undisguised assumption — declared and documented, not derived.

Data sources

Known limits

• The two real observation anchors (Maine, New Hampshire) are both New England — validation against them is a sanity check, not a real out-of-sample error estimate.
• A county-level price is, for the large majority of counties, a state or purely structural number applied uniformly across every county in that state or region — it is not a county-specific appraisal.
• Firewood price varies by species, seasoning, delivery distance, and season in ways this model does not observe.