In my last Blog entry I emphasized that
- The traditional DFA controls the time-shift by overemphasizing that part of the revision error variance which is due to phase artifacts of the asymmetric filter. This interpretation is simple, clear and straightforward – it is the only (known to me) formal a… pproach to disentangle amplitude and phase functions – but analytically intractable.
- The analytical DFA proposed in my previous two entries (with R-code!) tackles the time-shift problem by overemphasizing the imaginary part of the filter. This is a proxy. It cannot disentangle amplitude and phase functions properly. But it is tractable analytically.
Let me propose an alternative here which is appealing/efficient and tractable analytically but slightly less flexible because it emphasizes the transfer function i.e. the time-shift as well as the amplitude.