Université Paris 6
Pierre et Marie Curie | Université Paris 7
Denis Diderot | |

CNRS U.M.R. 7599
| ||

``Probabilités et Modèles Aléatoires''
| ||

**Auteur(s): **

**Code(s) de Classification MSC:**

- 65Cxx Probabilistic methods, simulation and stochastic differential equations {For theoretical aspects, see 68U20 and 60H35}
- 60J60 Diffusion processes [See also 58J65]
- 60-08 Computational methods (not classified at a more specific level) [See also 65C50]

**Résumé:** We are interested in approximating a multidimensional Itô process $(X_t)_{t\geq 0}$ killed when it leaves a smooth domain $D$: when the exit time is discretized along a regular mesh with time step $h$, we prove under a non characteristic boundary condition, that the discretization error is bounded from above by $C_1\sqrt h$, extending a previous result Gobet(\textit{Stoch. Proc. App. 2000}) obtained in the Markovian case under uniform ellipticity assumptions.
In the case of hypoelliptic diffusion processes and when a discrete Euler scheme is additionally used as an approximation of $X$, we prove that the upper bound for the weak error is still valid and that a lower bound with the same rate $\sqrt h$ holds true, thus proving that the order of convergence is exactly $\frac 12$.
This provides a theoretical explanation of the well-known bias that we can numerically observe in that kind of procedure.

**Mots Clés:** *Weak approximation ; Killed processes ; Discrete exit time ; Overshoot above the boundary*

**Date:** 2003-03-13

**Prépublication numéro:** *PMA-807*