Time Series Analysis
author:zoxiii
【参考文献】王燕. 应用时间序列分析-第5版[M]. 中国人民大学出版社, 2019.
E ( x t ) = μ t , ∀ t ∈ T E(x_t)=\mu_t~~,~~\forall t \in T E(xt)=μt , ∀t∈T
D X t = γ ( t , t ) = γ ( 0 ) , ∀ t ∈ T DX_t=\gamma(t,t)=\gamma(0),\forall t \in T DXt=γ(t,t)=γ(0),∀t∈T
γ ( k ) = γ ( t , t + k ) \gamma(k)=\gamma(t,t+k) γ(k)=γ(t,t+k)
估计值:
γ ^ ( 0 ) = ∑ t = 1 n ( x t − x ‾ ) 2 n − 1 \hat \gamma(0)=\frac{\sum_{t=1}^{n}{(x_t-\overline x)^2}}{n-1} γ^(0)=n−1∑t=1n(xt−x)2
ρ k = γ ( t , t + k ) D X t ⋅ D X t + k = γ ( k ) σ x 2 = γ ( k ) γ ( 0 ) \rho_k=\frac{\gamma(t,t+k)}{\sqrt{DX_t·DX_{t+k}}}=\frac{\gamma(k)}{\sigma_x^2}=\frac{\gamma(k)}{\gamma(0)} ρk=DXt⋅DXt+kγ(t,t+k)=σx2γ(k)=γ(0)γ(k)
当 k ≪ n k\ll n k≪n时:
ρ ^ k ≈ ∑ t = 1 n − k ( x t − x ‾ ) ( x t + k − x ‾ ) ∑ t = 1 n ( x t − x ‾ ) 2 , ∀ 0 ≤ k ≤ n \hat \rho_k \approx \frac{\sum_{t=1}^{n-k}{(x_t-\overline x)(x_{t+k}-\overline x)}}{\sum_{t=1}^n(x_t-\overline x)^2}~~,~~\forall 0 \le k \le n ρ^k≈∑t=1n(xt−x)2∑t=1n−k(xt−x)(xt+k−x) , ∀0≤k≤n
在一个常数附近随机波动,而且波动的范围有界,无明显趋势及周期特征
2倍标准差公式
L B = n ( n + 2 ) ∑ i = 1 k ( ρ ^ i 2 n − i ) LB = n(n+2)\sum_{i=1}^{k}{(\cfrac{\hat \rho_i^2}{n-i})} LB=n(n+2)i=1∑k(n−iρ^i2)
其中
n为序列观察期数;k为指定延迟期数