半参非中的t-计分-华东师范大学.pdfVIP

3 ( ) No. 3 2011 5 Journal of East China Normal University (Natural Science) May 2011 : 1000-5641(2011)03-0001-11 t- 1 2 1 , , (1. , 200092; 2. , 200062) : t- EM , . EM Q , . , t-. : ; ; t-; Cook ; ; P-; EM : O212.2 : A DOI: 10.3969/j.issn.1000-5641.2011.03.001 t-type estimators and influence diagnostics in semiparametric nonlinear models 1 2 1 JIANG Rong , SHAO Ming-jiang , QIAN Wei-min (1. Department of Mathematics, Tongji University, Shanghai 200092, China; 2. Department of Mathematics, East China Normal University, Shanghai 200062, China) Abstract: t-type regression estimators for the semiparametric nonlinear models were defined, and the EM algorithm for the t-type estimators were given. Moreover, the consis- tency of t-type regression estimators was obtained. And on the basis of the Q-function in the EM algorithm, statistical diagnostics methods were investigated. Finally, the proposed algorithm was tested by simulation analysis and by a real data sample. Key words: semiparametric regression; nonlinear model; t-type regression estimator; Cook distance; local influence; penalized spline; EM algorithm 0 : Y = f (X , β) + g (t ) + ε , i = 1, , n, (0.1) i i i i f (, ) β ; g () [0,1] p , ; Xi = (xi1 , , xip ) ∈ R ti ∈ [0, 1] p ; β = (