Under the trajectory law the actions are independent Bernoulli variables
ProvedTPG.integral_fun_actionscausal-inferencemarkov-chainnonstationaritypolicy-gradienttreatment-effect
Setting. is a finite state space, actions are , time is (the source's time ). The environment has transition matrices , initial distribution and reward laws on . For a sequence of treatment probabilities, and denote the law of the trajectory when , the actions are independent with , , and the rewards are conditionally independent with (Env.trajLaw T q); the Bernoulli policy is .
Lemma. Let for all . For every function of the action path ,
So are independent with , whatever the transition kernels are. The lemma is used to compute the moments of the inverse-propensity scores in the variance bounds.
Preamble
import Mathlib import Definitions.Def_TPG_Model import Definitions.Def_TPG_Markov open MeasureTheory ProbabilityTheory open scoped BigOperators ENNReal
Formal statement
namespace TPG
theorem integral_fun_actions {S : Type} [Fintype S] [MeasurableSpace S] [MeasurableSingletonClass S]
(E : Env S) (T : ℕ) (q : ℕ → ℝ) (hq : ∀ t : ℕ, 0 ≤ q t ∧ q t ≤ 1)
(g : (Fin T → Bool) → ℝ) :
∫ ω, g (fun t => Z t ω) ∂(E.trajLaw T q)
= ∑ z : Fin T → Bool, (∏ t : Fin T, if z t then q (t : ℕ) else 1 - q (t : ℕ)) * g z := 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 expresses the Bernoulli randomization design of §2.3, pp. 9–10 (the policy π_θ assigns treatment by i.i.d. Bernoulli(θ) variables, independently of all other randomness), for a time-varying sequence of probabilities.