I have to agree with Bill, MESA seems a better vehicle to investigate. I am not sure of how useful it can be, but I beleive it's worth investigating.
FFT on the other hand seems to have some basic conceptual issues that can not be overcome. Love to hear more discussion. ----- Original Message ----- From: wavemechanic To: [email protected] Sent: Tuesday, March 27, 2007 6:00 AM Subject: Re: [amibroker] FFT No, I'm not saying that MESA will give better results than a "better" FFT (is MESA a "better" FFT?). That judgment cannot be made until you leave the hypothetical and have a "better" FFT to talk about. Until then statistics help identify valid cycles and MESA offers some advantages, including noise filtering and ability to handle shorter cycles. Good luck in your search. Bill ----- Original Message ----- From: Ton Sieverding To: [email protected] Sent: Tuesday, March 27, 2007 3:24 AM Subject: Re: [amibroker] FFT So what you are saying is - 'Beyond that one can go to MESA' - that even after I should have found whatever modified version of FFT, MESA will give me better results. In other words, why playing with FFT if MESA is the right way to go. Is that your opinion or am I missing something ? Ton ----- Original Message ----- From: wavemechanic To: [email protected] Sent: Monday, March 26, 2007 2:42 PM Subject: Re: [amibroker] FFT The restrictions associated with FFT that Ehlers mentions can be found in any textbook. As for better results with FFT, the next step is to evaluate the cycles statistically (e.g., Bartels, F-ratio, chi-square, etc.). Beyond that one can go to MESA and such. Bill ----- Original Message ----- From: Ton Sieverding To: [email protected] Sent: Monday, March 26, 2007 2:58 AM Subject: Re: [amibroker] FFT Frankly for me these are John Ehlers typical arguments to use his MESA model in stead of FFT and has nothing to do with a discussion. The question for me still remains if there really is no way to get better results with FFT than the ones we have got ? If Fourier analysis is correct and it's possible to simulate whatever continues timeseries with a bunch of sinewaves and if MESA can give me the correct harmonics, it should also be possible to obtain the same results with a modified version of FFT. Question is how ? Ton Sieverding. ----- Original Message ----- From: wavemechanic To: AmiBroker, User Sent: Monday, March 26, 2007 1:27 AM Subject: Re: [amibroker] FFT There is a discussion of FFT use and problems on Ehlers MESA website: http://www.mesasoftware.com/fftcomparison.htm Bill ----- Original Message ----- From: Ara Kaloustian To: AB-Main Sent: Sunday, March 25, 2007 3:16 PM Subject: [amibroker] FFT I was playing with AB's FFT code that TJ provided... The cycles seem to shift relative to the data, based on how many data points are analyzed. This is of course expected. Question: Has anyone found a way to determine optimum number of data points to analyze, and then determine the relevance of the dominant cycle, or find any relevant cycles? Most of the time the dominant cycle seems to be the largest one available. Has anyone been able to use these cycles succesfully? Ara -------------------------------------------------------------------- No virus found in this incoming message. Checked by AVG Free Edition. Version: 7.5.446 / Virus Database: 268.18.17/732 - Release Date: 3/24/2007 4:36 PM ------------------------------------------------------------------------ No virus found in this incoming message. Checked by AVG Free Edition. Version: 7.5.446 / Virus Database: 268.18.18/733 - Release Date: 3/25/2007 11:07 AM ---------------------------------------------------------------------------- No virus found in this incoming message. Checked by AVG Free Edition. Version: 7.5.446 / Virus Database: 268.18.18/734 - Release Date: 3/26/2007 2:31 PM
