Phenomenological Structure for the Large Deviation Principle in Time-Series Statistics