Dislikedmbkennel,
I tend to agree with a lot of your assessments, and find your background very interesting....Ignored
QuoteDislikedNevertheless, I'll be the first to admit I'm probably the only poster that doesn't have metaquotes, so hopefully I'll be able to add something useful regarding your code. As limited as my understanding is, I looked over your code briefly, and as I interpret it, you are essentially conditioning some of the HULL parameters adaptively as a function of a local SNR value....
Compare to what you would get for a random walk.
QuoteDislikedSignal is usually extracted as the power of dominant component of the spectrum. That is the common signal processing approach at least. Your approach, If I'm understanding, sounds a bit like fractal type interpretation (like hurst, D, or R/S range type measurement). I don't know, there are a few ways to express this, but let me know if I'm on the right page first.
QuoteDislikedDid you come up with these adaptive equations on your own, or is there some reference to where they are shown elsewhere (for instance, using hyperbolic tangent as part of the scaling param).
QuoteDislikedDo you expect strong autocorrelation in SNR?
If there isn't then there's nothing useful to do.
Firstly it's very well known that returns in financial series are strongly heteroskedastic, meaning that the first order model of price is a random walk with a variance parameter which fluctuates itself with time. High volatility is more likely to be followed by high volatility and vice versa. Also, financial time series are of course causal in contrast to some models, news releases are preceded by low volatility followed by high-but-decaying volatility---the time-reversal does not happen. (conventional stochastic random walk processes are statistically time-reversible after detrending, nonlinear dynamics is not).
QuoteDislikedAlso, I like the idea about multifractals requiring multiple time frames to express, as it was mentioned earlier in this thread, but I'm not quite sure how you incorporate that concept here.
Thanks,
Jam
P.S. Regarding the power weighted kernel function, could you write out a simple explanation of what that is doing and why it's useful? thanks.
QuoteDislikedP.S.S. As of yet, no one has expressed whether or not any of these functions can be reduced or expressed as a simple linear weighted coefficient FIR.
The powerlaw kernel is nothing but FIR filter---the difference being that correlations (up to the cutoff) decay more slowly than typical filters used in conventional signal processing which always want to band-pass.
As I said, this is just raw material, but an unexplored direction. I'm here to see what this stimulates in others. I don't have answers.