Linearization / delta method for Markov chain statistics (Lemma EC.5)
DisprovedMarkovChainCLT.delta_method_of_uniformly_ergodicLet be a Harris ergodic, uniformly ergodic Markov chain with kernel and stationary distribution , started from . Let be a finite-dimensional statistic with measurable, square-integrable coordinates, let be its vector of coordinatewise sample averages and its stationary mean, and let be (Fréchet) differentiable at with derivative . Then
where is the asymptotic variance (autocovariance series) of the linearized scalar observable . This packages the delta method (Step 2 of Appendix EC.3.1 of arXiv:2407.19618) together with the linearization Lemma EC.5 (adapted from Lemma 6 of Farias et al. 2022): the asymptotic distribution of a differentiable functional of chain statistics is the CLT of its linearization, with variance .
RETIRED (deprecated by the mission captain). This statement is not provable as written, and is false for a suitable estimator. TendstoInDistribution carries an almost-everywhere measurability obligation on the estimator composed with the empirical statistics, and Frechet-differentiability at the single point where the derivative is taken does not supply it: with non-measurable is differentiable at with derivative and is measurable nowhere else, while the empirical statistics have a non-atomic law whenever the rewards do. The source is not at fault: arXiv:2407.19618 Definition 1 speaks of differentiable estimators and Appendix EC.3.2 uses " is differentiable (and thus continuous)", so the intended estimator class is continuous, hence measurable; the hypothesis was lost in formalization. Replaced by MarkovChainCLT.delta_method_of_uniformly_ergodic_of_measurable, which is identical apart from the added measurability hypothesis.
import Definitions.Def_MarkovAsymptoticVariance import Definitions.Def_MarkovErgodicity import Definitions.Def_MarkovChainPathMeasure open MeasureTheory ProbabilityTheory Filter open scoped NNReal ENNReal Topology
theorem MarkovChainCLT.delta_method_of_uniformly_ergodic {X : Type*} [MeasurableSpace X]
(P : Kernel X X) [IsMarkovKernel P] (π : Measure X) [IsProbabilityMeasure π]
(hP : MarkovChainCLT.HarrisErgodic P π)
(huni : MarkovChainCLT.UniformlyErgodic P π)
{ι : Type*} [Fintype ι]
(u : X → ι → ℝ) (hu : ∀ p, Measurable (fun x => u x p))
(hL2 : ∀ p, MemLp (fun x => u x p) 2 π)
(g : (ι → ℝ) → ℝ) (g' : (ι → ℝ) →L[ℝ] ℝ)
(hg : HasFDerivAt g g' (fun p => ∫ x, u x p ∂π)) :
TendstoInDistribution
(fun (n : ℕ) (ω : ℕ → X) =>
Real.sqrt n * (g (fun p => MarkovChainCLT.sampleAvg (fun x => u x p) n ω)
- g (fun p => ∫ x, u x p ∂π)))
atTop (id : ℝ → ℝ) (fun _ => MarkovChainCLT.chainMeasure P π)
(gaussianReal 0
(MarkovChainCLT.asymptoticVariance P π (fun x => g' (u x))).toNNReal) := by sorryRead-back
What the Lean code literally says, in plain math · claude-fable-5
Read-back: MarkovChainCLT.delta_method_of_uniformly_ergodic
Setting and binders. The theorem quantifies over:
- a type equipped with a measurable-space structure;
- a Markov transition kernel on (a kernel from to , with the typeclass assumption that every is a probability measure);
- a measure on , assumed (typeclass) to be a probability measure;
- a finite index type ;
- a map , i.e. a finite family of real-valued functions for ;
- a function on the finite-dimensional vector space ;
- a map that is a continuous linear map (this is the type of ; linearity and continuity are part of the data, not separate hypotheses).
Hypotheses.
-
Harris ergodicity (
hP), a custom definition unfolding to the conjunction:- is invariant under (in the Mathlib sense ), and
- for every starting point (not merely almost every), as .
Here is the custom iterated kernel: is the identity (Dirac) kernel , and (kernel composition). The distance is a custom total-variation distance defined as the supremum (real-valued
sSup) of the set , where are converted from extended-nonnegative reals to reals (an infinite measure value converts to ; for the probability measures appearing here all values are finite). This is the unnormalized sup-over-sets convention, taking values in for probability measures. -
Uniform ergodicity (
huni), a custom definition unfolding to: there exist real constants and with such that
with the same custom and as above. (Nothing is required at .)
-
Measurability (
hu): for every , the coordinate function is measurable. -
Square-integrability (
hL2): for every , the function lies in (Mathlib'sMemLpwith exponent : a.e.-strong measurability plus finite -norm). -
Differentiability (
hg): has as its Fréchet derivative at the single point whose -th coordinate is the Bochner integral . That is, differentiability of is assumed only at the mean vector of under (with derivative the given continuous linear map ), nowhere else; no continuity of the derivative in a neighborhood is assumed. (If some failed to be -integrable its Bochner integral would default to , moving the point of differentiation; under the probability-measure and hypotheses the integrals are genuine.)
Conclusion. Define, for a trajectory and , the sample-average vector with coordinates
i.e. the average of over states — the initial state is excluded, and for the value is (the reciprocal times an empty sum).
The theorem asserts convergence in distribution (Mathlib's TendstoInDistribution, i.e. weak convergence of the laws) of the real-valued random variables
as , where for every the underlying probability space is the path space equipped with the same measure : the custom chainMeasure , the law of the trajectory of the time-homogeneous Markov chain with transition kernel and initial distribution itself (a stationary start — the conclusion is stated only for this one initial distribution, built via the platform's Ionescu–Tulcea path-measure construction). The limit is expressed as the identity map on carrying the measure
i.e. the real Gaussian with mean and variance the nonnegative truncation (Real.toNNReal, i.e. ) of the custom asymptotic variance of the scalar function
the given linear functional applied to the vector . (When the variance parameter is , gaussianReal is the point mass at .)
The asymptotic variance unfolded. For , the custom lag- covariance is
with the iterated kernel above (so is the ordinary covariance of and under ). The asymptotic variance used in the conclusion is
i.e. , where each infinite sum is Lean's tsum: if the family is not summable, each sum silently evaluates to and the "asymptotic variance" degenerates to the plain stationary variance ; no summability hypothesis appears in the theorem. All integrals are Bochner integrals, defaulting to when the integrand is not integrable.
Edge cases silently included. The finite index type may be empty (then is a one-point space and is effectively a constant); the truncation means a negative value of the custom asymptotic-variance expression yields the degenerate limit ; and the conclusion concerns only the stationary initial distribution , not arbitrary starting laws or starting points.
Confirmed by the mission captain (proposal self-audit).