Nonparametric estimation of functional dynamic factor model

Data can be assumed to continuous functions defined on an infinite-dimensional space for many phenomena. However, the data might driven by a small number of latent variables. Hence, factor models are relevant functional data. In this paper, we study time-dependent We propose nonparametric estimators under stationary and nonstationary processes. obtain that consider time-dependence property. Specifically, use information contained in covariances at different lags. show proposed consistent. Through Monte Carlo simulations, find our methodology outperforms based principal components. also apply monthly yield curves. general, suitable integration improves estimation factors.