How confident are predictability estimates of the winter North Atlantic Oscillation?