The Setting
Finite amplification compares weighted moments of normalized finite sums with Gaussian moments. The weights here are fixed tent functions and selector weights, and the gain is a function of their moments of orders zero, one and two.
Let be the standard normal probability law. For a measurable space , a measure on it, a real observable and a family of real weights indexed by a set , the weighted moment array is defined by (1). (1)✦ The weighted means and weighted variances of a moment array are given by (2), with nonzero zeroth moments where the quotients are used: (2)
The Fixed Gaussian Weights
The clipping function and, for real parameters , the tent function are defined in (3). (3) For a grid half-size and spacing , let , whose middle index is . The grid centers , ramp differences and selector weights are the functions in (4): (4) The leaf index set is for the fixed grid parameters in (5), where is the grid radius and is the leaf tent width: (5) For each leaf index , the leaf test and selector test are the functions defined in (6): (6)
The moment gain of two arrays on a finite index set is the expression in (7): (7)✦ The Gaussian leaf and selector moment arrays and the gain threshold are defined by (8); denotes the identity observable on : (8) For observables on probability spaces , write and .
Finite Sums and Soft Events
For a finite set and a function , the uniform mean is the real number defined in (9); for nonempty , its uniform probability measure assigns every measurable subset the mass shown there: (9) For and , the normalized sum is defined in (10); averages on product sets are uniform: (10)✦
For a real observable on a finite set and weights , the finite weighted moment array is defined by (11): (11)✦ The soft event weight associated with families is the function in (12): (12) For a real observable on a finite set , a real shift and a weight , the quadratic deficit is the real number given by (13): (13) For the normalized sum, let and . The finite leaf moment array , shift and soft event weight associated with are given by (14): (14)
The Claim
Theorem 1. Let converge in distribution under to the standard normal law . For every continuous compactly supported function , the compact-test limit is (15): (15)✦ For every and leaf index , the leaf moment limit is (16): (16)✦
Suppose additionally that every is measurable, that is integrable for , and that for each such . For every and , the selector moment limit is (17): (17)✦
On any finite index set, let be moment arrays converging entrywise to , respectively, with at every index. Their gain limit is (18): (18)✦
Under the additional moment hypotheses on , there is an index for which (19) holds: (19)✦
Let be finite and nonempty, and let satisfy and . Its normalized sums have the moment limits (20), for : (20)✦ There is a positive integer for which the finite soft gain satisfies (21): (21)✦
In the paper, the result reads: if centered variance-one variables converge in distribution to a standard normal with , then for fixed , an -tuple of independent copies and a standard Gaussian -tuple satisfy for every measurable whose discontinuities are Gaussian-null and for which with a fixed constant .
What Is Checked
The transfer statements concern observables on probability spaces and the fixed Gaussian weights, with the additional integrability and moment convergence hypotheses for selector tests, while (20) and (21) concern centered variance-one observables on finite uniform spaces.