Finite Impulse Response Filter
by ANTHONY WARREN, Ph.D. with JACK K. HUTSON
Figure 1 and 2 show a comparison of computer output using Fourier analysis Power and then Amplitude
plots. It can be seen that the Power Spectrum plot, which is Amplitude squared, makes it very easy to
discern which cycles are significant and which are not. The Power Spectrum suggests that cycles with
periods of less than 20 days may not be important, in the long move. Figure 3 is an example of computer
output plotted from the following Finite Impulse Response (FIR) Filter subroutine using a Hanning
weighted moving average. A 41 day Hanning weight was used based on the 20 day interpretation from a
Fourier Power Spectrum (i.e. N = 2M + 1 e.g. 2(20) + 1 = 41 days).
The subroutine may be expanded to incorporate any number of moving average weighting schemes
between statements 1210 and 1230. This routine has been written to run using any computer that supports
the BASIC language and in addition a few extraneous variables (T$, P(?)) have been added to improve
graphic output if one is using the Compu Trac programming subsystem.