A chain started from its invariant law is strictly stationary
ProvedMarkovChainCLT.isStrictlyStationary_coord_chainMeasureLet be a Markov transition kernel on and let be an invariant probability distribution for , i.e. . Consider the law of the trajectory of the time-homogeneous Markov chain with kernel started from . Then the coordinate process is strictly stationary: for every , the shifted trajectory has the same law on the path space as .
This is the classical fact that a Markov chain started in its stationary distribution is a stationary process. It follows from invariance by induction on the finite-dimensional distributions: for every because is invariant, and the conditional law of given is the same kernel-generated law for every by time-homogeneity; the two together determine the law of the shifted path.
Every mixing central limit theorem used in this mission (Theorems 3 and 5-8) assumes a strictly stationary sequence, so this lemma is the gateway that lets those theorems be applied to a Markov chain run from equilibrium.
import Definitions.Def_MarkovErgodicity import Definitions.Def_MarkovChainPathMeasure import Definitions.Def_MixingCoefficients open MeasureTheory ProbabilityTheory Filter open scoped ENNReal NNReal Topology ProbabilityTheory
theorem MarkovChainCLT.isStrictlyStationary_coord_chainMeasure {X : Type*} [MeasurableSpace X]
(P : Kernel X X) [IsMarkovKernel P] (π : Measure X) [IsProbabilityMeasure π]
(hinv : Kernel.Invariant P π) :
IsStrictlyStationary (chainMeasure P π) (fun i (ω : ℕ → X) => ω i) := by sorry