Model selection with data-oriented penalty |
| |
| Affiliation: | 1. Department of Applied Mathematics, National Sun Yat-sen University, Taiwan;2. Department of Statistics, Penn State University, 326 Thomas Building, University Park, PA 16802-2111, USA;3. Department of Mathematics and Statistics, York University, Canada;1. Consorzio RFX, Corso Stati Uniti 4, 35127 Padova, Italy;2. Università degli studi di Padova, Via VIII Febbraio, 35122 Padova, Italy;3. Ecole Polytechnique Fédérale de Lausanne (EPFL), Swiss Plasma Center (SPC), CH-1015 Lausanne, Switzerland;4. CEA, IRFM, 13108 Saint-Paul-lez-Durance, France;5. Max-Planck-Institute for Plasma Physics, Boltzmannstr. 2, D-85748 Garching, Germany;1. University of Bradford School of Management, Emm Lane Campus, Bradford, West Yorkshire BD9 4JL, UK;2. The ICMA Centre, Henley Business School, University of Reading, PO Box 242, Reading RG6 6BA, UK;1. Institute for Biosecurity, Saint Louis University, College for Public Health & Social Justice, St Louis, MO;2. Department of Biostatistics, Saint Louis University, College for Public Health & Social Justice, St Louis, MO;3. School of Nursing, Saint Louis University, St Louis, MO;1. General Atomics, San Diego, CA, United States;2. PPPL, Princeton, NJ, United States |
| |
| Abstract: | We consider the problem of model (or variable) selection in the classical regression model using the GIC (general information criterion). In this method the maximum likelihood is used with a penalty function denoted by Cn, depending on the sample size n and chosen to ensure consistency in the selection of the true model. There are various choices of Cn suggested in the literature on model selection. In this paper we show that a particular choice of Cn based on observed data, which makes it random, preserves the consistency property and provides improved performance over a fixed choice of Cn. |
| |
| Keywords: | |
| 本文献已被 ScienceDirect 等数据库收录！ |
|
