Non and semi-parametric estimation in models with unknown smoothness Yulia Kotlyarova Dalhousie University Victoria Zinde-Walsh Department of Economics, McGill University 855 Sherbrooke Street West, Montreal, Quebec, Canada, H3A 2T7 tel. (514) 398 4834; fax (514) 398 4839 victoria.zinde-walsh@mcgill.ca April 28, 2006 Abstract Many asymptotic results for kernel-based estimators were estab- lished under some smoothness assumption on density. For cases where smoothness assumptions that are used to derive unbiasedness or as- ymptotic rate may not hold we propose a combined estimator that could lead to the best available rate without knowledge of density smoothness. A Monte Carlo example conrms good performance of the combined estimator. JEL code C14 Key words: nonparametric estimation, combined estimator Support of Social Sciences and Humanities Research Council of Canada (SSHRC) and of Fonds quØbecois de la recherche sur la sociØtØ et la culture (FRQSC) is acknowledged. We thank Marcia Schafgans for valuable comments. 1