In this article, we investigate the properties of the EBIC in variableselection for generalized linear models with non-canonical links and divergingnumber of parameters in ultra-high dimensional feature space. The selectionconsistency of the EBIC in this situation is established under moderateconditions. The finite sample performance of the EBIC coupled with a forwardselection procedure is demonstrated through simulation studies and a real dataanalysis.
展开▼