Convolution Function in DTSimulator

DTSimulator’s newest version now includes convolution. This means that it can now create powerful functions for frequency/severity models.

For example, if we have an operational risk with a mean frequency of 4, we could use a Poisson distribution to model how many times the event happens probabilistically. On the other hand, the severity is modeled by a LogNormal distribution with mean = 100 and standard deviation = 300. If we incorrectly multiply frequency times severity, the Poisson times the Lognormal, we would be generating n number of events at each iteration all of them with the same severity.

This multiplication would be overestimating the tails of the distribution.The correct way is with a convolution: we generate n number of events with the Poisson frequency distribution and then, for each one of these events, we generate different severities. This is convolution. It is integrated in a single powerful DTSimulator function. This guarantees a correct dimensioning of the compounded risk.