Time: 14.30-15.30 p.m., June 5th, 2019, Wednesday
Venue: Room 602, College of Mathematics, Mingxiang Campus
Speaker: Weixing Song, Kansas State University
Abstract:
Due to its conceptual simplicity, Simulation-Extrapolation (SIMEX) enjoys great popularity in measurement error modeling in reducing the bias caused by the measurement errors. However, the traditional SIMEX approach can be very computationally extensive in implementation. In this talk, an accelerated version of SIMEX algorithm is proposed. By applying the conditional expectation step directly to the target function, the proposed algorithm successfully removes the actual simulation step for a variety of statistical models, thus significantly reduces the computational time. It is noted that in some cases, the proposed method provides the exact form of the extrapolation function, and the resulting estimate can be obtained by simply setting the extrapolation variable to -1. In general, extrapolation is still necessary. Several examples are used to illustrate the effectiveness of the proposed methods.
From College of Mathematics and College of Data Science
Go Back