A0151
Title: Multiple testing boosting for spatial panel and regression models
Authors: George Kapetanios - Kings College London (United Kingdom) [presenting]
Abstract: High-dimensional model selection remains a well-researched yet challenging problem in econometrics and statistics. A variant of a sequential stagewise model selection procedure is developed, incorporating a parsimonious multiple testing stopping rule. This combination appears to be relatively novel. The approach is applied to the problem of determining network structures for spatial panel models. Several useful theoretical characteristics are derived, and the procedure exhibits desirable small-sample properties. Furthermore, the procedure is generalized to regression models, where optimality results are established, and its excellent, relatively small-sample performance is again demonstrated.
