Assumption 1 contracts signed measures of total mass zero:
ProvedTPG.mixing_contractioncausal-inferencemarkov-chainnonstationaritypolicy-gradienttreatment-effect
Let the transition matrices of the environment satisfy Assumption 1 on the horizon with a constant : for every , every action and every pair of probability vectors on the finite state space , , with .
Lemma. Let for all , let , and let be a real row vector on with . Then
This is the form in which the mixing assumption enters every bias, variance and mixing bound of the source.
Preamble
import Mathlib import Definitions.Def_TPG_Model import Definitions.Def_TPG_Estimators import Definitions.Def_TPG_Markov import Definitions.Def_TPG_Bridge open MeasureTheory ProbabilityTheory open scoped BigOperators ENNReal
Formal statement
namespace TPG
theorem mixing_contraction {S : Type} [Fintype S] [DecidableEq S]
(E : Env S) (T : ℕ) (γ : ℝ) (hγ : 0 ≤ γ) (hmix : E.Mixing T γ)
(q : ℕ → ℝ) (hq : ∀ t : ℕ, 0 ≤ q t ∧ q t ≤ 1) (a b : ℕ) (hab : a ≤ b) (hb : b ≤ T)
(μ : S → ℝ) (hμ : ∑ x, μ x = 0) :
∑ x, |Matrix.vecMul μ (((List.Ico a b).map (fun s => E.PMix (q s) s)).prod) x|
≤ γ ^ (b - a) * ∑ x, |μ x| := by sorry
end TPGSource
Johari, Peng, Xing, Estimation of Treatment Effects Under Nonstationarity via the Truncated Policy Gradient Estimator, revised manuscript of October 2026 (revision of arXiv:2506.05308v3, https://arxiv.org/abs/2506.05308); numbering follows the revised manuscript. Auxiliary statement, not a numbered result of the paper: it is eq. (24), p. 38 (Appendix F.1): by Jordan decomposition the contraction of Assumption 1 extends to signed measures of total mass zero and to multi-step kernels.