FIRf.afl & FIRf.png 8/11/06 Dave Howarter- FIR Filter Cycles,
Fred and I have been working on improving the FIR filter cycling. (Well 80% Fred and 20% me
).
The attached version has two significant improvements. -- The prediction beyond the end of data has been improved to try to match a cosine as closely as possible to the recent data. This is now a white wave projecting beyond the last data.
-- If you set an end-pole with <shift><F12> at some date location, the calculation of optimium cycle lengths will be made as if that data was the last day of data and the projection of data beyond that will be made from that data. This can be used to test "OOS" and see how well it would have predicted the future data.
-- a measure of the cycle amplitude is now displayed.
I think the measure of amplitude is important information and we are working on ideas to use it to help predict the future movement. There can be some shifting of the FIR output that can cause the exact timing of peaks and valleys to be off, but I think the FIR creates a pretty accurate measurement of the cycle amplitude. Of course this changes from cycle to cycle, so we list the average of the last 3 cycles. Hurst has a basic principle that the amplitude of the wave doubles every time the period doubles. But the FIR filter gives very different results, which I trust. The results For 8/11/06 are: