Pilot CEO and innovation: A model averaging approachfor divergent-dimensional Poisson regressions
报告人： 张新雨，中科院数学与系统科学研究院/预测中心研究员
时间：2020-06-08 10:00-11:00
地点：讲座视频链接： https://pan.baidu.com/s/1w5lHOzixGbSUb891axDsJQ 提取码: t2pu（七天内有效）
Abstract:
Poisson regression is the benchmark model to analyze count dependent variables. This paper proposes a new model averaging method to address model uncertainty in Poisson regressions, allowing the dimension of covariates to increase with the sample size. We utilize an unbiased estimator of the Kullback-Leibler (KL) distance to choose weights for combining candidate models. We show that when all candidate models are misspecified, the proposed model averaging estimate is asymptotically optimal in the sense that it yields a KL loss that is asymptotically identical to that resulting from the infeasible best possible averaging estimator. In a different situation where there exist correct models among the candidate models, our model averaging estimates can also produce consistent estimates of the coefficients. We apply the proposed techniques to study the relation between pilot CEO and corporate innovation outcome, and our estimates suggest that this relation is less robust when taking into account the large degree of model uncertainty typically faced by corporate financial economists.
About the Speaker:
张新雨，中科院数学与系统科学研究院/预测中心研究员，国家杰出青年基金获得者。在中科院系统所获得博士学位学位，曾在TAMU做博士后研究。主要从事计量经济学和统计学的理论和应用研究工作，具体研究方向包括模型平均、机器学习和组合预测等。担任期刊《Journal of Systems Science and Complexity》领域主编、期刊《SADM》、《系统科学与数学》等的AE或编委，是双法学会数据科学分会副理事长和国际统计学会当选会员。
Place: 腾讯会议
会议 ID：966 568 452
会议密码：123123
会议链接：https://meeting.tencent.com/s/OwF9MUUWedAY
