Structural economic models are getting increasingly complex. This complicates their implementation. In particular, for many models, (i) identification of the relevant components from data is still unresolved; (ii) estimation and testing procedures tend to be somewhat ad hoc and with no theoretical underpinning; (iii) the implementation of model and estimators often require numerical approximations since model solutions are not available on closed form.The proposed research aims at addressing these three issues, (i)-(iii), by developing new methods and results for identification, estimation, testing and implementation of structural models. Particular emphasis is put on the use of nonparametric techniques. Many of the projects involve empirical applications where the proposed methods will be taken to data.The individual projects making up the proposal are: Identification and inference in discrete choice models; identification and inference in continuous choice models; Implementation and estimation of parametric structural models; Filtering and likelihood inference in dynamic latent variable models; Bandwidth selection in estimation and testing of non- and semiparametric models.