Answers are grounded in this paper and cite their sources.
In the paper spherically invariant processes are defined. The characteristic function of these processes (ξ (t)) in accordance with Shonberg’s theorem [1] is of the form \[ χ (η ) ≡ { M}eiη = f({ D}η ) = ∫_0^∞ {e^{ - γ { D}η } } G(dγ ), \]η = ∫ ξ (t)η (t)dt, where G is some measure on [ 0,∞ ). Only if the process is spherically invariant, then 1) every extrapolation problem has a linear solution, 2) every functional transformation leaving the correlation function of the process invariant retains its measure in the space of realizations. If a spherically invariant process is stationary and ergodic, then it is Gaussian.
No takes yet. Share an insight, caveat, or question.
A. M. Vershik (1964) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: