Abstract:We study the problem of nonparametric estimation of the linear multiplier function $\theta(t)$ for processes satisfying stochastic differential equations of the type $$dX_t= \theta(t)X_t dt+ \epsilon\; \sigma_1(t,X_t)\sigma_2(t,Y_t)dW_t, X_0=x_0, 0 \leq t \leq T$$ where $\{W_t, t\geq 0\}$ is a standard Brownian motion, $\{Y_t, t\geq 0\}$ is a process adapted to the filtration generated by the Brownian motion. We study the problem of estimation of the unknown function $\theta(.)$ as $\epsilon \rightarrow 0$ based on the observation of the process $\{X_t,0\leq t \leq T\}.$
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the application of non - parametric estimation in stochastic differential equations, especially the non - parametric estimation of the linear multiplier function \(\theta(t)\) with multiplicative stochastic volatility. Specifically, the paper studies the stochastic differential equation in the following form:
\[ dX_t=\theta(t)X_t dt+\epsilon\sigma_1(t, X_t)\sigma_2(t, Y_t)dW_t, \quad X_0 = x_0, \quad 0\leq t\leq T \]
where:
- \(\{W_t, t\geq0\}\) is a standard Brownian motion;
- \(\{Y_t, t\geq0\}\) is a stochastic process adapted to the filtration generated by the Brownian motion;
- \(\theta(t)\) is an unknown linear multiplier function;
- \(\epsilon\) is a small positive parameter.
The main objective of the paper is to perform non - parametric estimation of the unknown function \(\theta(t)\) when \(\epsilon\rightarrow0\) under the observation of the process \(\{X_t, 0\leq t\leq T\}\), and to study the properties of the estimator.
### Main contributions
1. **Non - parametric estimation method**: The paper proposes a non - parametric estimator \(\hat{\theta}_t\) based on the kernel method for estimating the linear multiplier function \(\theta(t)\).
2. **Consistency and asymptotic distribution**: The paper proves the consistency and asymptotic distribution properties of the estimator \(\hat{\theta}_t\), especially its performance when \(\epsilon\rightarrow0\).
3. **Theoretical results**: The paper provides several theorems, which describe in detail the convergence rate and asymptotic distribution of the estimator under different conditions.
### Key assumptions
- The function \(\theta(t)\) is bounded on the interval \([0, T]\).
- The functions \(\sigma_1(t, x)\) and \(\sigma_2(t, x)\) satisfy certain growth and boundedness conditions.
- The process \(\{X_t, 0\leq t\leq T\}\) has a unique strong solution.
### Theorem overview
- **Theorem 3.1**: Under certain conditions, the estimator \(\hat{\theta}_t\) is consistent.
- **Theorem 3.2**: Under more stringent conditions, the mean - square error of the estimator \(\hat{\theta}_t\) has a specific asymptotic behavior.
- **Theorem 3.3**: Gives the asymptotic distribution of the estimator \(\hat{\theta}_t\), especially the asymptotic normality when \(\epsilon\rightarrow0\).
### Application background
This type of problem has important applications in fields such as financial engineering, physics, and biology, especially when dealing with systems with stochastic volatility. Through non - parametric estimation methods, complex stochastic processes can be handled more flexibly without having to assume a specific functional form in advance.