This talk covers important recent advances in
analyzing the stochasticity of stationary and non-stationary
time series. We give an overview on the Markovian
methods  for analyzing stochastic data in
time and scale by inference of an evolution equation.
Based on this reconstruction of the underlying stochastic
process, we nvestigate the cascade of information
from large scales to small scales (related to intermittency
of the time series), multiscale correlation functions and
level crossing properties. We use the method to investigate complex
fluctuations in heart interbeats, financial time series,
rough surfaces, turbulence, and seismic time series.