博士生讨论班2024[12]
时间：2024-12-06
阅读量：73次
- 演讲人： 程思昂
- 时间：2024年12月10日14:00
- 地点：浙江大学紫金港校区行政楼1417报告厅
报告文章：M-estimator and model selection for hidden Markov models (Siang Cheng)
摘要：M-estimators are widely used in various statistical inference problems due to their excellent properties. The most well-known among them is MLE, derived from KL divergence. We adopt the similar approach to get a new M-estimator from a different distance measure between distributions called Renyi’s pseudodistance. The corresponding estimator is called minimum Renyi’s pseudodistance estimator. Then we extend the MRPE to Markov models. Theoretical analysis shows that the MRPE has consistency and asymptotic normality. Moreover, our simulations indicate that it is more robust than MLE.