A measurable functional generates a smaller -algebra
ProvedMarkovChainCLT.processSigma_comp_lemeasure-theoryprobability
Let be a family of maps into a measurable space and let be measurable. Then for every index set ,
Indeed is the comap of the target -algebra along , which factors as the comap along of the comap along ; measurability of says the latter is contained in the -algebra of , and comap is monotone.
This is the elementary fact underlying the statement that the mixing coefficients of a functional process are dominated by those of the underlying process — the supremum defining the functional coefficient ranges over a subfamily of the pairs of events used for the original one.
Preamble
import Definitions.Def_MixingCoefficients open MeasureTheory ProbabilityTheory Filter open scoped ENNReal NNReal Topology ProbabilityTheory
Formal statement
theorem MarkovChainCLT.processSigma_comp_le {Ω X E : Type*} [MeasurableSpace X]
[MeasurableSpace E] (Y : ℕ → Ω → X) (g : X → E) (hg : Measurable g) (s : Set ℕ) :
processSigma (fun i ω => g (Y i ω)) s ≤ processSigma Y s := by sorrySource
G. L. Jones, "On the Markov Chain Central Limit Theorem", Probability Surveys 1 (2004) 299-320, arXiv math/0409112v2, Section 4, proof of Corollary 1 (arXiv v2 p. 10): "Let alpha(n) and alpha_f(n) denote the strong mixing coefficients for the Markov chain X = {X_n} and the functional process {f(X_n)}, respectively. By an earlier remark alpha_f(n) <= alpha(n) for all n >= 1."