1. INTRODUCTION 5
for the increment-tamed Euler-Maruyama scheme (1.5). Moreover, Corollary 2.27
in Subsection 2.3.1 yields uniform moment bounds for the fully drift-implicit Euler
scheme and Lemma 2.28 in Subsection 2.3.2 establishes uniform moment bounds
for partially drift-implicit approximation schemes. These results on uniform mo-
ment bounds are applications of a general theory which we develop in Section 2.1.
In this theory (see Propositions 2.1 and 2.7 and Corollaries 2.2, 2.3 and 2.6) we
assume a Lyapunov-type inequality to be satisfied by the approximation processes
on large subevents of the probability space, i.e., on complements of rare events; see
inequality (2.11) in Corollary 2.2. One of our main results (Theorem 2.13 in Sub-
section 2.2.1) establishes this Lyapunov-type condition for the Euler-Maruyama
approximations (1.4). More precisely, whereas the Euler-Maruyama approxima-
tions often do not satisfy a Lyapunov-type inequality on events of probability one
in the case of superlinearly growing coefficients according to Corollary 2.17 in Sub-
section 2.2.1, the Euler-Maruyama approximations do satisfy the Lyapunov-type
inequality (2.11) on large subevents of the probability space according to Theo-
rem 2.13 in Subsection 2.2.1. This integrability result on the Euler-Maruyama
approximation processes can then be transfered to a large class of other one-step
approximation processes. To be more precise, Lemma 2.18 in Subsection 2.2.2
proves that if two general one-step approximation schemes are close to each other
in the sense of (2.102) (see Lemma 2.18 for the details) and if one approximation
scheme satisfies the Lyapunov-type inequality (2.11) on large subevents, then the
other approximation scheme satisfies the Lyapunov-type inequality (2.11) on large
subevents of the probability space as well. After having established the Lyapunov-
type inequality (2.11) on such complements of rare events, the general rare event
based theory in Section 2.1 can be applied to derive moment bounds and further
integrability properties of the approximation processes. In Chapter 3, we then pro-
ceed to study convergence in probability (see Section 3.3), strong convergence (see
Section 3.4) and weak convergence (see Section 3.5) of approximation processes for
SDEs. Definition 3.1 in Section 3.2 specifies a local consistency condition on ap-
proximation schemes which is, according to Theorem 3.3 in Section 3.3, sufficient for
convergence in probability of the approximation processes to the exact solution of
the SDE (1.1). This convergence in probability and the uniform moment bounds in
Corollary 2.21 then result in the strong convergence (1.7) of the increment-tamed
Euler-Maruyama approximations (1.5); see Theorem 3.15 in Subsection 3.4.3 for
the details. Moreover, we obtain results for approximating moments and more gen-
eral statistical quantities of solutions of SDEs of the form (1.1) in Section 3.5. In
particular, Corollary 3.23 in Subsection 3.5.2 establishes convergence of the Monte
Carlo Euler approximations for SDEs of the form (1.1).
1.1. Notation
Throughout this article, the following notation is used. For a set Ω, a measur-
able space (E, E) and a mapping Y : Ω E we denote by σΩ(Y ) := {Y
−1(A)

Ω: A E} the smallest sigma algebra with respect to which Y : Ω E is measur-
able. Furthermore, for a topological space (E, E) we denote by B(E) := σE(E) the
Borel sigma-algebra of (E, E). Moreover, for a natural number d N and two sets
Previous Page Next Page