Compound Poisson simulation

benchmarks several versions of code for generating Compound Poisson random variables

Description

Compound Poisson random variables are of the form S=X1+...+XN where the Xj are iid random variables and N is random with the Poisson distribution. Compound Poisson variables have many applications in physics and finance.

This file benchmarks several codes for generating 1 million random samples from a Compound Poisson distribution where the terms are Lognormal with mu=0 and sigma=1 and the Poisson frequency is 10.

The run times vary by a factor of more than 200 between the fastest and slowest versions (in R2009b). Interestingly, the fastest code is 10 times faster than the neat 1-liner that uses ARRAYFUN.

Requires POISSRND (Statistics Toolbox) and RANDRAW (FEX #7309).

Acknowledgements

The author wishes to acknowledge the following in the creation of this submission:
RANDRAW

• Required Products: Statistics Toolbox
• MATLAB release: MATLAB 7.9 (2009b)