Abstract:We consider a simple mean reverting diffusion process, with piecewise constant drift and diffusion coefficients, discontinuous at a fixed threshold. We discuss estimation of drift and diffusion parameters from discrete observations of the process, with a generalized moment estimator and a maximum likelihood estimator. We develop the asymptotic theory of the estimators when the time horizon of the observations goes to infinity, considering both cases of a fixed time lag (low frequency) and a vanishing time lag (high frequency) between consecutive observations. In the setting of low frequency observations and infinite time horizon we also study the convergence of three local time estimators, that are already known to converge to the local time in the setting of high frequency observations and fixed time horizon. We find that these estimators can behave differently, depending on the assumptions on the time lag between observations.
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is to estimate the parameters and local time in a threshold diffusion model. Specifically, the paper focuses on a mean - reverting diffusion process with piecewise - constant drift and diffusion coefficients and is discontinuous at a fixed threshold. The paper discusses methods for estimating these parameters from discrete observational data, in particular, using the Generalized Method of Moments Estimator (GME) and the Maximum Likelihood Estimator (MLE) for estimation. In addition, the paper also studies the asymptotic theory of these estimators when the time span of observations tends to infinity in the cases of low - frequency and high - frequency observations.
### Background and Problem Description of the Paper
The paper considers a diffusion process \(X_t\) that satisfies the following stochastic differential equation (SDE):
\[ X_t = X_0+\int_{0}^{t}\sigma(X_s)dW_s+\int_{0}^{t}b(X_s)ds, \quad t\geq0, \]
where \(\sigma(x)\) and \(b(x)\) are piecewise - constant functions and may be discontinuous at some fixed threshold \(r\in\mathbb{R}\):
\[ \sigma(x)=\begin{cases}
\sigma_+ > 0 & \text{if }x\geq r,\\
\sigma_- > 0 & \text{if }x < r
\end{cases} \]
\[ b(x)=\begin{cases}
b_+\in\mathbb{R} & \text{if }x\geq r,\\
b_-\in\mathbb{R} & \text{if }x < r
\end{cases} \]
It is assumed that the initial condition \(X_0\) is deterministic and \(b_->0\) and \(b_+ < 0\) so that the process is mean - reverting and ergodic.
### Research Objectives
The main objective of the paper is to estimate the drift parameters \(b_+\) and \(b_-\), as well as the diffusion parameters \(\sigma_+\) and \(\sigma_-\) from discrete observational data. Specific methods include:
1. **Generalized Method of Moments Estimator (GME)**: Constructed based on discrete observational data, using the theoretical long - term behavior of conditional moments and local time.
2. **Maximum Likelihood Estimator (MLE)**: Assuming that the process is continuously monitored and the time span of observations tends to infinity.
### Asymptotic Theory
The paper develops the asymptotic theory of these estimators, considering two different asymptotic settings:
1. **Low - frequency observations**: The fixed time interval \(h\), and the number of observations \(N\) tends to infinity.
2. **High - frequency observations**: The time interval \(h_N\) tends to zero while the time span \(T_N = Nh_N\) tends to infinity.
### Local Time Estimation
The paper also studies the asymptotic behavior of three different local time estimators under low - frequency observations and long - time spans. These estimators are known to converge to the local time at a rate of \(N^{1/4}\) under high - frequency observations, but under low - frequency observations, their asymptotic behavior may be different, depending on the observation time interval \(h\).
### Related Work
The paper mentions that for diffusion processes with piecewise - linear drift (i.e., the threshold Ornstein - Uhlenbeck process), the parameter estimation problem has been widely studied. However, for diffusion processes with piecewise - constant drift, there are relatively few related studies.
### Conclusions
The paper provides the asymptotic properties of these estimators under different observation settings, proves their consistency and asymptotic normality, and shows that under high - frequency observations, the discrete maximum likelihood estimator (dMLE) converges to the maximum likelihood estimator (MLE) for continuous observations at a rate of \(N^{1/4}\). In addition, the paper also discusses the asymptotic behavior of local time estimators under low - frequency observations.