A stationary distribution of a finite chain vanishes off the recurrent states (p. 80)
ProvedYoungConventions.Perturbed.stationary_vanishes_off_recurrentmarkov-chainp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1recurrent-classstationary-distribution
Let be a Markov chain on a finite set , let be a stationary distribution of , and let be a state that is not recurrent under (some state reachable from cannot reach ). Then
In the proof of Theorem 4 this is applied to the limit and the unperturbed chain : the stochastically stable states lie in the recurrent classes of , so the potential only needs to be computed on recurrent states.
Formalization Note The page states the fact for and ; it is stated here for an arbitrary finite chain and an arbitrary stationary distribution.
Preamble
import Mathlib import Definitions.Def_YoungConventions_Perturbed_FiniteChain
Formal statement
namespace YoungConventions.Perturbed
/-- Stationary distributions vanish off the recurrent states (Young 1993, Econometrica 61:57–84,
Appendix, p. 80, PDF p. 25: "Since μ⁰ is a stationary distribution of P⁰, μ⁰_z = 0 for every
state z that is not recurrent under P⁰").
For every Markov chain `A` on a finite set `X`, every stationary distribution `ν` of `A` and
every state `z` that is not recurrent under `A`, `ν z = 0`.
**Formalization Note.** The page applies this to `μ⁰` and `P⁰`; it is stated here for an arbitrary
finite chain, which is the fact used. -/
theorem stationary_vanishes_off_recurrent {X : Type*} [Fintype X] [DecidableEq X]
(A : Matrix X X ℝ) (hA : A ∈ Matrix.rowStochastic ℝ X)
(ν : X → ℝ) (hν : IsStationaryDist A ν) (z : X) (hz : ¬ IsRecurrentState A z) :
ν z = 0 := by sorry
end YoungConventions.Perturbed
Source
Young (1993), The Evolution of Conventions, Econometrica 61:57–84, Appendix, p. 80 (PDF p. 25), after the proof of Lemma 1
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.