Dan Simpson • 6/14/2022

Tail stabilization of importance sampling etimators: A bit of theory

This technical article delves into the theory behind importance sampling, a Monte Carlo method for estimating expectations. It highlights the problem of unstable estimators caused by extreme importance ratios when proposal and target distributions differ. The core solution discussed is Truncated Importance Sampling (TIS), which caps large ratios with a threshold to stabilize variance, analyzing the inherent bias-variance trade-off introduced by this technique.

0 comments

#Computational Statistics #Monte Carlo #Probability Theory