Penalized Psuedo-Likelihood Estimation of General Conditional Single-Index Response Models


Ming-Yueh Huang
2009-10-23  12:30 - 14:30
Room 405, Mathematics Research Center Building (ori. New Math. Bldg.)

In this talk, we concern with an appealing single-index conditional density model with an unspecified response function. When the considered regression coefficients are sparse, a pseudo-likelihood estimation using the adaptively weighted L_1 penalty is proposed for the index coefficients. An extension of our approach to the high-dimensional predictors is further addressed. (This is a joint work with Chin-Tsang Chiang)