Discussion on “Is Average Run Length to False Alarm Always an Informative Criterion?” by Yajun Mei

Abstract

We comment on Mei's critique of the standard minimax formulation with the ARL to false alarm of a sequential detection rule as its operating characteristic, and provide an alternative Bayesian approach for his use of a mixing distribution to handle pre-change parameters. We also consider computational issues in implementing the ARL and better alternatives for evaluating sequential detection rules in complex stochastic systems.

Keywords:

• Bayesian change-point
• Hidden Markov models
• Importance sampling and resampling

Subject Classifications:

• 62L15
• 60G40
• 62F15

ACKNOWLEDGMENTS

Chan's research is supported by the National University of Singapore under grant R-155-000-062-112. Lai's research is supported by the National Science Foundation under grant DMS-0305749.

Notes

Recommended by Alex Tartakovsky