Equation (5), §3, p. 8 — credit assignment: the DN estimator equals (1/N) Σ_i (η_i + ξ_i Σ_{j∈N_i} η_j) Y_i
ProvedDiffNeighbors.Bias.credit_assignmentdifferences-in-neighborsestimator-identitynetwork-interferencep2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1
Consider units with interference graph (neighbourhoods ), outcome functions and observed outcomes at a treatment vector . For write
The DN estimator (3), , can be rewritten as assigning credit for each outcome to the treatments of and of its neighbours:
The paper uses this form in the analysis of the estimator's expectation and variance: each outcome enters once, with a weight built from the treatments that influence it.
Differences from the source:
- generalization: p ∈ (0,1) and outcomes under Assumption 1 (standing) → any real p and any outcome functions f, with no Assumption 1; reason: the identity is a rearrangement of a finite double sum that uses only the symmetry of the graph; it specializes to the paper's setting
- encoding: units [N] = {1,…,N}, treatment vectors z ∈ {0,1}^N, an undirected graph without self-loops, d = max_i |N_i| → a finite type ι with N = card ι, z : ι → Bool with 1/0 read through ind, G : SimpleGraph ι, N_i = G.neighborFinset i, d = G.maxDegree; reason: a simple graph is exactly an undirected graph without self-loops; same instances
Preamble
import Mathlib import Definitions.Def_DiffNeighbors_Bias_Model
Formal statement
namespace DiffNeighbors.Bias
open Finset
/-- **Credit assignment**, (5), §3, p. 8: the DN estimator (3) equals
`(1/N) ∑_i (η_i + ξ_i ∑_{j ∈ N_i} η_j) Y_i`, with `η_j = z_j/p − (1 − z_j)/(1 − p)`. -/
theorem credit_assignment {ι : Type*} [Fintype ι] [DecidableEq ι] (G : SimpleGraph ι)
[DecidableRel G.Adj] (p : ℝ) (f : ι → (ι → Bool) → ℝ) (z : ι → Bool) :
dnEstimator G p f z = (Fintype.card ι : ℝ)⁻¹ *
∑ i, (eta p z i + xi p z i * ∑ j ∈ G.neighborFinset i, eta p z j) * f i z := by sorry
end DiffNeighbors.Bias
Source
Peng, Ye, Zheng, Differences-in-Neighbors for Network Interference in Experiments, arXiv:2503.02271v1, p. 8, §3, Credit Assignment, (5)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.