Asymptotic Robustness of Estimators in Rare-Event Simulation (with P. L’Ecuyer, P. Glynn, and B. Tuffin).

Summary:

This paper studies a hierarchy of efficiency notions in rare event simulation. It does so in a wide range of environments, such as reliability and large deviations settings. Some of the results that I particularly find interesting in this paper are counter-examples showing that even if you approximate the zero-variance importance sampling distribution reasonably well you still can get a sub-optimal estimator. Another nice observation (in my view) relates to the implications of having an asymptotically vanishing relative error estimator (this yields total variation approximations).

Bibtex:

@Article{LecBlaTufGly09,

author = { P. L’Ecuyer and J. Blanchet and B. Tuffin and P. Glynn},

title = {Asymptotic robustness of estimators in rare-event simulation},

journal = {ACM Transactions on Modeling and Computer Simulation - TOMACS},

year = {2010},

volume = {20},

pages = {1-41}

}