Large deviations for scaled sums of \(p\)-adic-valued rotation-symmetric independent and identically distributed random variables (Q2181631)

The paper under review studies the asymptotic order of the logarithm of the mass of such sets off the origin for an appropriately scaled sum of \(p\)-adic valued, independent, identically and rotation-symmetrically distributed random variables. The author proves the large deviation principle with a rate function being constant except at the origin. The \(p\)-adic field \(\mathbb{Q}_p\) with the ultra-metric property carries a compact and open topology, and is a separable and complete metric space where many methods of stochastic analysis can be applied. For I.I.D. random variables \(\xi_i\) on the \(p\)-adic field, Proposition 1 of the article under review shows the asymptotic probability of the complement of any compact open set \(K\) in \(\mathbb{Q}_p\), \[ \mathop{\lim_{n\to \infty}}\nolimits^{\pm} \frac{1}{n}\log P_n (K^c) = -\alpha \theta_{\mp} \log p, \] where \(\lim^+ = \lim \sup\), \(\lim^- = \lim \inf\), \(P_n\) is the law of the scaled sum \(p^{n+c_n}\sum_{i=1}^{[N(n)]}\xi_i\) with the integer part of \(N(n) = \frac{p^{2\alpha}(p-1)}{p^{\alpha +1}-1} T_1(n)^{-1}\), \(T_1(n)\) measure the tail probability for a positive constant \(\alpha\) and a function \(L\) on \(\mathbb{Z}\) such that \(\lim_{n\to \infty} \frac{L(n+1)}{L(n)} = 1\), and \(\theta_{\pm} = \lim^{\pm} \frac{c_n}{n}\). Proposition 2 states that \[ \lim\nolimits^{\pm} \frac{\log P_n(B)}{n} \le - \alpha \theta_{\mp} \log p, \] for any closed set \(B\) in \(\mathbb{Q}_p\) not including the origin \[ \lim_{n\to \infty} \log P_n (A) =0 \text{ and} \] for any open set \(A\) in \(\mathbb{Q}_p\) including the origin. With these bounds, the main result, Theorem 1, follows by assuming proper conditions on \(c_n\) and the function \(L\), \[ \mathop{\lim_{n\to \infty}}\nolimits^{\pm} \frac{\log P_n(B(a, p^l))}{n} = - \alpha \theta_{\mp} \log p \] for any ball \( B(a, p^l)\) not including the origin. Section 2 prepares for technical Lemma 1 to understand an inductive relation of \(U_n(m) =P(|\sum_{i=1}^n\xi_i|=p^m)\) from the I.I.D. rotation-symmetrically random variables \(\xi_i\), and Proposition 3 to calculate the tail probabilities of the sum \(\xi_i\) based on the Lemma 1 and the \(T_1(m)\) information. The tail probability \(T_n(m)\) can be derived more precisely by using Fourier transform by using a result of the author from an earlier work [J. Theor. Probab. 13, No. 3, 635--657 (2000; Zbl 0971.60008)]. Section 3 is devoted to the proofs of Proposition 1, 2 and Theorem 1. Proposition 1 follows from a technical lemma 2 on the upper and lower bounds for the probability of the complement of the open ball, computing the asymptotic logarithm on both sides and then the upper and lower limits. Proposition 2 follows from Proposition 1 and Lemma 2. The main result, Theorem 1, follows from a rotational symmetry, which implies \[ P_n(B(a, p^l)) = \frac{P_n(B_{k-1}^c) - P_n(B_k^c)}{p^{k-l} - p^{k-l-1}}, \] and the upper bound from Lemma 2 with an independent property, and the straightforward limit of asymptotic terms. The article ends with the proof of Corollary 1 that specifies the rate function in the asymptotic behavior.