Abstract:We establish several features of the stationary solution of purely explosive autoregressions of order d based on nonstandard initial values.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the existence and properties of stationary solutions in purely explosive autoregressions. Specifically, the author studies the stationary solutions and their characteristics of the d - order purely explosive autoregressive model under non - standard initial values.
### Core Problems of the Paper
1. **Purely Explosive Autoregressive Model**: When the maximum absolute value of the characteristic root \(\rho(B)>1\), the model is considered to be purely explosive. In this case, traditional autoregressive models usually do not produce stationary solutions because their solutions will diverge over time.
2. **Existence of Stationary Solutions**: The author attempts to prove that under such purely explosive conditions, a stationary solution can still be obtained by choosing specific initial conditions. Specifically, the author proposes a special initial condition \(U_0 = -\sum_{j = 1}^{\infty}B^{-j}W_j\), so that the generated sequence \((U_n)_{n\geq0}\) is stationary and ergodic.
3. **Properties of Stationary Solutions**: After finding the stationary solutions, the author further explores the probability properties of these stationary solutions, including properties such as the central limit theorem and mixing convergence. These properties are very important for statistical inference and parameter estimation.
### Main Contributions
- **Existence of Stationary Solutions**: The author has proven that under specific initial conditions, the purely explosive autoregressive model can have stationary solutions.
- **Probability Properties**: The author has analyzed in detail the probability properties of stationary solutions, especially the central limit theorem and the mixing convergence theorem, which are crucial for understanding the behavior of the model and conducting statistical inferences.
- **Applications**: These results have important applications in fields such as financial time - series analysis, especially when dealing with non - causal autoregressive processes.
### Formula Summary
- Purely Explosive Condition: \(\rho(B)>1\), where \(\rho(B)\) is the spectral radius of matrix \(B\).
- Initial Condition for Stationary Solution: \(U_0 = -\sum_{j = 1}^{\infty}B^{-j}W_j\).
- Central Limit Theorem: For the stationary solution \(U_n\), we have
\[
\frac{1}{\sqrt{n}}\sum_{k = 1}^n U_k\xrightarrow{d}N\left(0,\sigma^2\left(\sum_{i = 1}^d\theta_i^{-1}\right)^2E_d\right)
\]
where \(E_d\) is a \(d\times d\) matrix with all elements being 1.
Through these works, the author provides a solid theoretical basis for understanding and dealing with purely explosive autoregressive models.