Precise Asymptotics in the Law of the Iterated Logarithm of Moving-Average Processes

Yun Xia Li,Li Xin Zhang

DOI: https://doi.org/10.1007/s10114-005-0542-4

2006-01-01

Abstract:In this paper, we discuss the moving-average process X_k = ∑_i = - ∞^∞α _i + kε _i , where α i ;-∞ < i < ∞ is a doubly infinite sequence of identically distributed φ-mixing or negatively associated random variables with mean zeros and finite variances, α i ;-∞ < i < ∞ is an absolutely summable sequence of real numbers. Set S_n = ∑_k = 1^n X_k ,n ⩾ 1 . Suppose that σ ^2 = Eε ^2_1 + 2∑_k = 2^∞Eε _1ε _k > 0 . We prove that for any δ⩾ 0, if E[ ε ^2_1( log log| ε _1|)^δ - 1] < ∞ , lim_∈↘ o∈ ^2δ + 2∑_n = 1^∞( log log n)^δ/n log nP{| S_n|⩾ετ√(2n log log n)} = 1/( δ + 1)√(π)Γ( δ + 3/2), , and if E[ ε ^2_1( log| ε _1|)^δ - 1] < ∞ , lim_∈↘ o∈ ^2δ + 2∑_n = 1^∞( log n)δ/nP{| S_n|⩾ετ√(n log n)} = μ ^( 2δ + 2)/δ + 1τ ^2δ + 2 , where τ = σ·∑_i = - ∞^∞α _i ,Γ( ·) is a Gamma function and μ (2δ+2) stands for the (2δ + 2)-th absolute moment of the standard normal distribution.