Theorem 2.4, p. 3 â for g â đ, (ÈČâ(1) â ÎŒ)/g(ÈČâ) â B(1)/g(B) under Assumption (2.1)
ProvedGlynnSTS.Limit.theorem_2_4Let be a real-valued measurable process, for , and suppose Assumption (2.1) holds: for finite constants and , in , where is a standard Brownian motion and . Let , i.e. is measurable and (i) for ; (ii) for ; (iii) ; (iv) . Then
This is display (2.5), the basic limit theorem of the method of standardized time series. The unknown variance constant does not appear in the limit, so quantiles of give asymptotic confidence intervals for the steady-state mean without estimating .
Formalization Note Convergence in distribution is Mathlib's TendstoInDistribution. Division by zero returns in Lean; this affects only the event and, in the limit, the null event , and matches the paper's convention for in the proof.
import Mathlib import Definitions.Def_GlynnSTS_Limit_Setting open MeasureTheory ProbabilityTheory Filter Topology open scoped NNReal ENNReal
namespace GlynnSTS.Limit
theorem theorem_2_4 {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
(B : Ω â C(unitInterval, â)) (hB : IsStdBMC P B)
(Y : â â Ω â â) (Ybar : â â Ω â C(unitInterval, â)) (ÎŒ Ï : â)
(h21 : Assumption21 P Y Ybar ÎŒ Ï B)
(g : C(unitInterval, â) â â) (hg : ClassM P B g) :
TendstoInDistribution (fun (n : â) Ï => (Ybar n Ï 1 - ÎŒ) / g (Ybar n Ï)) atTop
(fun Ï => B Ï 1 / g (B Ï)) (fun _ => P) P := by sorry
end GlynnSTS.Limit
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.