1: \begin{abstract}

2: We provide a unified framework to obtain numerically certain quantities,

3: such as the distribution function, absolute moments and prices of

4: financial options, from the characteristic function of some (unknown)

5: probability density function using the Fourier-cosine expansion (COS)

6: method. The classical COS method is numerically very efficient in

7: one-dimension, but it cannot deal very well with certain integrands

8: in general dimensions. Therefore, we introduce the damped COS method,

9: which can handle a large class of integrands very efficiently. We

10: prove the convergence of the (damped) COS method and study its order

11: of convergence. The method converges exponentially if the characteristic

12: function decays exponentially. To apply the (damped) COS method, one

13: has to specify two parameters: a truncation range for the multivariate

14: density and the number of terms to approximate the truncated density

15: by a cosine series. We provide an explicit formula for the truncation

16: range and an implicit formula for the number of terms. Numerical experiments

17: up to five dimensions confirm the theoretical results.

18: 

19: \textbf{Keywords: }Fourier-transform, numerical integration, inversion

20: theorem, COS method, CDF, rainbow options\\

21:  \textbf{Mathematics Subject Classification} 65D30 · 65T40 · 60E10

22: 62P05

23: \end{abstract}

24: