A diagnostic test for identifying nonlinear dynamical relationships in
time series, based on mutual information and redundancy, functionals
introduced in information theory, is proposed. Its ability to distingu
ish (noised) multiperiodic and random time series from time series gen
erated by chaotic dynamical systems is demonstrated. The latter are ch
aracterized by specific behaviour of marginal redundancies reflecting
the increase of uncertainty in time due to positive information produc
tion rate.