Multidimensional large deviation local limit theorems (Q1081959)

This paper presents a large deviation local limit theorem for arbitrary multidimensional random variables. More specifically let \(\{T_ n\}\) be a sequence of random variables with \(f_ n\) being the density function of \(T_ n/n\). Conditions are imposed on the sequence of moment generating functions of \(\{T_ n\}\) which allow an asymptotic expression for \(f_ n(m_ n)\) to be derived, provided the sequence \(\{m_ n\}\) satisfies an additional condition. The asymptotic expression involves the large deviation rate Chernoff function of \(T_ n/n\). When \(T_ n\) is the sum of i.i.d. random variables the corresponding results of \textit{W. Richter} [Teor. Veroyatn. Primen. 3, 107-114 (1958; Zbl 0080.344)] are shown to be consequences of the results here. The authors considered the one dimensional case in an earlier paper [Ann. Probab. 13, 97-114 (1985; Zbl 0559.60030)].