Should Behaviouristic Equation in Time Series be Computed in Terms of Absolute Level Variables or their Deltas?