1: \begin{abstract}

2: We analyse a multilevel Monte Carlo method for the approximation 

3: of distribution functions of univariate random variables.

4: Since, by assumption, the target distribution is not known

5: explicitly, approximations have to be used. 

6: We provide an asymptotic analysis of the error and the cost 

7: of the algorithm. 

8: Furthermore we construct an adaptive version of the algorithm

9: that does not require any a priori knowledge on weak or strong

10: convergence rates. We apply the adaptive algorithm

11: to smooth path-independent and path-dependent functionals and 

12: to stopped exit times of SDEs.

13: \end{abstract}

14: