Moderate deviations via cumulants (Q354766): Difference between revisions

Moderate deviation principles are established for a large class of random variables satisfying certain bounds on the cumulants. The proof utilizes a lemma from \textit{R. Rudzkis, L. Saulis} and \textit{V. Statulyavichus} [Lith. Math. J. 18, 226--238 (1978; Zbl 0423.60027)]. The general theorem enables results to be established for certain dependent random variables as well as new results for partial sums of independent but not necessarily identically distributed random variables. The majority of the paper is devoted to novel applications of the general result. Applications to dependency graphs and subgraph counting statistics in Erdős-Rényi random graphs complement moderate deviation results obtained in [\textit{H. Döring} and \textit{P. Eichelsbacher}, Electron. J. Probab. 14, 2636--2656 (2009; Zbl 1193.60032)]. Other applications include moderate deviation results for non-degenerate \(U\)-statistics and, in random matrix theory, moderate deviation principles are established for the characteristic polynomials for the circular unitary ensemble (CUE), circular orthogonal ensemble (COE) and circular symplectic ensemble (CSE). The final application is to the number of particles in a growing box of random determinantal point processes such as the number of eigenvalues in a Gaussian unitary ensemble.