Importance Sampling (Rare-Event Variance Reduction)
Quant ResearchLevel: Intermediate
Importance Sampling (Rare-Event Variance Reduction)
Hints
- What is the estimator and which random variables produce it?
- Separate bias, variance and computational cost; how would you check convergence?
Answer
To estimate E_p[f(X)], sample from q and average f(X)p(X)/q(X). q must be positive wherever f·p contributes, and variance analysis needs a finite second moment of the weighted payoff. Shift sampling toward important rare events without creating extreme weights. Forgetting the likelihood ratio changes the target expectation. Examine weighted-payoff variance and independent replications; an apparently large weight-based effective sample size alone does not certify accurate rare-event estimation.
Source and licence: Quantitative Finance: From Mathematical Foundations to Trading Desk Reality — zcoderz · MIT · answer: Quantoros editorial