The Setting
An observation of a sign cube retains a conditional sum of coordinates. Grouping independent observations places these conditional scores on disjoint coordinate blocks. Let be a finite coordinate set, and write for its sign cube with the uniform law. An observation is a function that reports a value in a set .
For a real-valued function on a finite set , the uniform mean is the average in (1), with value zero on the empty set.✦ (1) Here and in the scalar normalizations below, . For nonempty , its uniform probability measure assigns the masses in (2) to measurable subsets of .✦ (2)
The cell indicator of an observation at a value is the indicator of the fiber ; its cell probability is its uniform mean, as in (3). (3) A cell degree bound for an observation is a common bound on the ordinary real multilinear Fourier degrees of its cell indicators , for , with constants and the zero function assigned degree zero.
For an observation , a value and a function , the restricted moment is the real number given by the uniform mean of , and the fiber conditional mean is the real number obtained by dividing by the cell probability . The conditional average of a function given an observation evaluates the fiber conditional mean at , as in (4).✦ (4) The fiber conditional mean is zero when the cell probability is zero.
For an observation , the coordinate sum is the function that sums the coordinates, and its conditional score is the conditional average of this sum given , as in (5). The retained variance of an observation is the second moment of its centered conditional score under the uniform law on .✦ (5) For an observation and a value , the standardized conditional value is the real number obtained by dividing the fiber conditional coordinate sum by . The standardized score of an observation evaluates the standardized conditional value at , as in (6).✦ (6)
For a finite block index set , the block coordinate map is the bijection in (7).✦ The grouped observation applies to each block.✦ (7)
Write for a nonnegative integer , with . For a function and a nonnegative integer , the normalized finite sum is the function in (8).✦ (8) For an observation and a nonnegative integer , the grouped score scale is the real number obtained by multiplying the square roots of the block count and the retained variance . The normalized grouped score is the normalized finite sum of the standardized block scores, as in (9).✦ (9)
The Claim
For a probability measure and a real-valued random variable , write for the distribution of under , and for weak convergence of probability measures. The standard normal law on is denoted by .
Theorem 1. Let be a probability measure on a measurable space , and let be measurable with and . Under the product laws , (10)✦ For a nonempty finite set and a function with and , the uniform laws satisfy (11)✦ Let be an observation. For every , its normalized grouped score has mean (12)✦ If and , its grouped conditional score satisfies (13)✦ and the normalized second moment is (14)✦
In the paper, the result reads: for and a cell degree bound , independent copies of the conditional score give , the joint observation has cell degree bound and score , and converges in distribution to .
What Is Checked
The limit statements (10) and (11) concern product probability measures and uniform laws on finite Cartesian powers; the score identities (12), (13) and (14) concern observations on finite sign cubes, with positive retained variance and a positive block count required for the latter two identities.