Simulation

[EarlGrey]

Are there any good papers or other sources that anyone could recomend for simulation?

Also, are there professional standards which come presenting results to clients? The reason I ask is that recently I did a simulation in @Risk of ground up losses. The Actuary I work with chose a distribution which did not conform to the empirical data whatsoever. The reason for choosing the distribution was so we could simulate ground up losses to get certain average values in several excess layers.

I just wonder if chossing a distribution that does not conform to your empirical data is right.

[Kitten]

I just wonder if chossing a distribution that does not conform to your empirical data is right.

It may or may not be right...

In your situation, I would myself ask the following questions:
1. Is the data credible?
2. Have I considered the impact of outliers?
3. How good of a fit am I getting from various statistical distributions (I use BestFit, which adds in to @Risk)?
4. Finally, does it pass the "smell test" (does it make sense)? :p

The answers to the above questions will impact any distributions used in any @Risk model I would build. I might have missed some questions, but those were the first that sprang to mind. You have to be willing to justify at least the above questions to yourself and others when asked if you are going to do this type of work, I would think. I'm not sure of any specific standards covering this off, though.

Hope that helps. :)

[Sherwin]

I often use the simulation to price a XOL treaty. As my experience, I think that you could get an idea about the scope of the parameters of the frequently used loss model. Just for example, the parameter alpha of Pareto Distribution should be in the interval (0.6, 2). If the modeled parameter is 3.2, you could have an idea that the data is not quite credible or other thing is wrong. At that time, we may use the industry data instead of the company's own data after a careful data review perhaps.