Smooth regression finds causal process.

Natural processes are smooth.
Noise is not.

The Gamma test uses this difference
to identify causal connections in data
and measure predictive strength. It is
an error estimator for all smooth process,
"the holy grail of nonlinear modeling".

New Tools in Nonlinear Modeling and Prediction
A paper describing the Gamma test and its use.

R package

or text to 541-490-7291