.png){kind=icon}
.png){kind=icon}
Return
On robust estimation for moderate deviations from a unit root
.png){kind=icon}
PRE
AI
DOI：10.1007/s11749-026-01020-z.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
We in this paper propose a robust estimation technique for a first-order autoregressiveprocess characterized by the root rho(n )= 1 + c/k(n), employing a kernel mode-basedobjective function construed on the mode value. Compared to traditional least squaresor maximum likelihood approaches, the newly developed method exhibits enhancedrobustness against outliers and non-normal errors. We suggest a computationallyefficient mode expectation-maximization algorithm leveraging a Gaussian kernel tonumerically estimate the coefficient. Under mild assumptions, we derive the asymp-totic distributions of the resulting kernel mode-based estimator, assuming thatknincreases to infinity at a slower rate thann. Specifically, for c < 0, we establish aconvergence rate of root nk(n )with a normal limit distribution, while for c > 0, the con-vergence rate is k(n)rho(n )(n)with a Cauchy limit distribution. Monte Carlo simulations arepresented to illustrate the favorable finite sample performance of the proposed esti-mation procedure. Furthermore, we extend these results to the general autoregressiveprocess with a coefficient satisfying n|1-rho(n)|-> infinity under weaker initial conditions.The convergence rates are demonstrated to be[n(1-rho(2)(n))(-1)](1/2 )and rho(n)(n)/(rho(2)(n )- 1)for nearly stationary and mildly explosive cases, respectively.
Keywords:
Kernel
Mildly explosive
Mode
Moderate deviations
Nearly stationary
Unit root