Having multivariate time series, e.g., financial or economic
data observed at regular time intervals, we want to describe
the components of the time series with a smaller number of
uncorrelated factors.
The usual factor model of multivariate analysis cannot be
applied immediately as the factor process also varies in time.
There is a dynamic part, added to the usual factor model, the
auto-regressive process of the factors.
Dynamic factors can be identified with some latent driving
forces of the whole process. Factors can be identified only by
the expert (e.g., monetary factors) .