Drift estimator (source code)

= Drift estimator
{title2=$\widehat b_T(x,h)$}

For continuous observations of an <Itô diffusion> $dX_t=b(X_t)dt+\sigma(X_t)dW_t$, a local drift estimator at $x$ is
$$
\widehat b_T(x,h)=\frac{\int_0^T\mathbf1_{\{|X_t-x|\leq h\}}dX_t}{A_T},\qquad A_T=\int_0^T\mathbf1_{\{|X_t-x|\leq h\}}dt,
$$
on $\{A_T>0\}$. Its error is the average of $b(X_t)-b(x)$ over that window, plus $A_T^{-1}\int_0^T\mathbf1_{\{|X_t-x|\leq h\}}\sigma(X_t)dW_t$. A local <Hölder continuity> bound with exponent $\alpha$ bounds the first term by $Lh^\alpha$.

The commonly used noise scale $(Th)^{-1/2}$ requires occupation control. For a stationary model with continuous positive density $\pi(x)$ of its <invariant distribution> and locally bounded $\sigma$, the <Itô isometry> makes the numerator's second moment $O(Th)$. If additionally $A_T/(2Th\pi(x))\to1$ in <probability> in the chosen long-time, shrinking-window regime, division gives stochastic error $O_{\mathbb P}((Th)^{-1/2})$. Positivity of the invariant density alone does not establish that concentration of the random denominator.