Generalized dynamic factor models are multivariate time series models. A high dimensional stochastic process can be representated as a sum of a latent and an error process which are uncorrelated. The latent process is driven by dynamic factors. The error process is not assumed to be iid in the cross-section as in the Frisch factor models, but is allowed to be weakly correlated in the cross-section and over time.
Forni, Hallin, Lippi, Reichlin and Stock, Watson developed estimation procedures, which are presented and compared.