Special moments (Q2483916)

This article is devoted to the investigation of the problem of matching the first \(n\) moments of a random variable \(Y\) with a linear combination of independent copies of another random variable \(X\). The author proves the following result: A linear combination \(Z\) of \(n\) independent, unbiased Bernoulli random variables \(X_1, X_2, ..., X_n\) can match the first \(2n\) moments of a random variable \(Y\) which is uniform on an interval. More generally, for each \(p \geq 2,\) each \(X_k,\;k = 1, 2, ..., n,\) can be uniform on an arithmetic progression of length \(p\). All the values of \(Z\) lie in the range of \(Y,\) and their ordering as real numbers coincides with the dictionary order on \((X_1, X_2, ..., X_n).\) The construction involves the roots of truncated \(q\)-exponential series. The precise mathematical form of these results is summarized by the following Theorem and Conjecture that are proved, respectively enounced, in the paper: Theorem: Let \(p \geq 2\) be an integer, let \(X\) be a uniform random variable on the set \(\{p - 1,\;p - 3,\) \(p - 5,\;...,\;1 - p\}\), and let \(X_1, X_2, ..., X_n\) be independent copies of \(X\). Then there exist unique constants \(a_1 > a_2 > ... > a_n > 0\) such that the first \(2n\) moments of the linear combination \(Z = a_1X_1 + a_2X_2 +\cdots+ a_nX_n\) agree with the first \(2n\) moments of a random variable \(Y\) which is uniform on \([- 1, 1].\) Moreover, the following inequality holds: IneqTheor: \[ \sum^{n}_{j=1}| a_{j} - p^{-j} | < 1/[p^{n}(p-1)]. \] Conjecture: The enouncement is identical to the Theorem above, except that the IneqTheor is strengthened to the following: IneqConject: \[ \sum^{n}_{j=1}(p_j -1) | a_j - \prod^{j}_{k=1} p_k^{-1} | < \prod^{n}_{k=1} p_k^{-1}. \] These results are useful for higher-dimensional cubature problems based on error-correcting codes, in particular extended BCH codes, to obtain a (\(2n + 1)\)-cubature formula on a \(k\)-cube \([ - 1, 1]^k\) with equal weights and \(O(k^n)\) points.