Almost sure convergence of the Bartlett estimator (Q2368601)

Suppose \( \{ Y_k \}\) is a fourth order stationary sequence with \(EY_i = 0\) and \( \gamma_j = \text{Cov}(Y_0,Y_j)\). Denote by \(s^2_n\) the Bartlett (= variance) estimator and put \(\sigma^2 = \sum_j \gamma_j\). The authors establish the almost sure consistency \(s^2_n \rightarrow \sigma^2\) in the both the weak dependence and long memory case. Their conditions involve fourth order cumulants and assumptions on the rate of increase of the sum in the \(s^2_n\) definition.