Telescoping estimate: for factors of modulus
ProvedMartingale.norm_prod_sub_prod_le_sumLet be complex numbers of modulus at most . Then for every ,
The proof is a telescoping induction. Writing and ,
and since (a product of factors of modulus ) and , the triangle inequality bounds the left side by , which the inductive hypothesis closes.
Why the hypothesis is exactly modulus . Without it the estimate is false — replacing one factor in a product of large numbers amplifies the perturbation by the size of the remaining factors. Boundedness by is what keeps the amplification factor at most at each step, so the errors merely add rather than compounding.
Where it is used. This is the aggregation step of McLeish's proof of the martingale central limit theorem. The individual comparison is a statement about one factor; this lemma converts a family of such per-factor estimates into a bound on the difference of the two products, turning into control of . Combined with the negligibility of the increments and the boundedness of the sum of squares, that is what drives .
The statement is elementary and entirely general — no probability enters — so it is reusable wherever finite products of bounded complex numbers are compared.
import Mathlib.Analysis.SpecialFunctions.Complex.Circle open Finset
theorem Martingale.norm_prod_sub_prod_le_sum (a b : ℕ → ℂ) (ha : ∀ k, ‖a k‖ ≤ 1) (hb : ∀ k, ‖b k‖ ≤ 1) (n : ℕ) :
‖∏ k ∈ Finset.range n, a k - ∏ k ∈ Finset.range n, b k‖
≤ ∑ k ∈ Finset.range n, ‖a k - b k‖ := by sorry