CMStatistics 2016: Start Registration
View Submission - CFE
A1745
Title: Nonparametric estimation of time-varying parameters in nonlinear models Authors:  Young Jun Lee - University College London (United Kingdom) [presenting]
Dennis Kristensen - University College London (United Kingdom)
Abstract: Asymptotic properties are developed for nonparametric estimators of time-varying parameters for a general class of dynamic models. We take as given that the parameters of interest are identified as the maximizer of some population moment. The time-varying parameters are then estimated using local versions of standard M-estimators based on this population moment. Under high-level conditions including local stationarity, we show that the estimators are consistent and asymptotically normally distributed. We provide primitive conditions for our high-level conditions for Markov models. Our methodology and theory generalize existing nonparametric methods for time-varying parameters. To demonstrate the usefulness of our general set-up, we give explicit conditions to hold the established asymptotic properties for several specific nonlinear time series models.