A special feature of Probability and Statistics Day at UMBC 2014 is that the conference, including the workshop, is open to all statistics graduate students from UMBC and local universites free of charge; however, REGISTRATION IS REQUIRED! The deadline to register is Friday, April 11, 2014. // REGISTER NOW
For more information, contact any member of the organizing committee:
Bimal Sinha
Conference Chair
443.538.3012
Kofi Adragni
410.455.2406
Yvonne Huang
410.455.2422
Yaakov Malinovsky
410.455.2968
Thomas Mathew
410.455.2418
Nagaraj Neerchal
410.455.2437
DoHwan Park
410.455.2408
Junyong Park
410.455.2407
Anindya Roy
410.455.2435
Elizabeth Stanwyck
410.455.5731
Participant Information
Priyam Mitra
Paper: Frequentist Model Averaging : A General Framework and Theories
In this paper, we propose a general framework for frequentist model averaging and study the asymptotic behavior of such model estimators. The proposed framework broadens the scope of existing methodology by including models even with large biases. We are interested in the problem of estimating unknown model parameters as well as prediction. When the true model is unknown, the traditional approach is to select a `right' model among a set of candidates and further analysis proceeds as though the selected model was the true model that was known apriori. This approach ignores the uncertainty introduced in the process of model selection. A model averaging method addresses this issue by combining estimators of the parameter by averaging them for a set of candidates so that it incorporates the underlying model uncertainty. Assuming the data are from an unknown model, we derive the model average estimator and describe the limiting distributions and risk properties while taking potential modelling bias into account. We use a selection of weights to combine the individual estimators with the weights chosen so that they minimize an unbiased estimate of the mean square error of the model average estimator. To demonstrate this idea we use a linear regression framework and combine estimators from a set of candidate models and discuss its asymptotic properties. A simulation study is performed to compare the performance of the estimator with that of existing methods. The results show the benefits of incorporating multiple models in the estimator rather than a single one.