Bob Fulks' Excellent Digital Filter Illumination

[rbl@xxxxxxxxxxxxxxxxx (Robin B. Lake)]

Re: Digital Filters

Following up on Bob Fulks' excellent explaination of digital filters ...

Digital filters can be described most easily as the sequence of weights
that Bob presented.  However, it is also possible to describe these
filters:
-  In terms of their impulse response:  the sequence of numbers produced
        by the filter when the filter is applied to a 0 1 0 0 0 ... 0 0
        sequence (of stock prices);
-  In terms of their step response:  the sequence of numbers produced
        by the filter when the filter is applied to a 0 1 1 1 1 1 ... 1 1
        sequence (of stock prices);
-  In terms of a Fourier Transform of the filter's weights, but this
        produces complex-valued numbers, such a 3.123 + j0.43 .

I believe testing any and all filters with a step and an inpulse of
stock prices is simple due dilligence for any trader.  When you know not
what is in the "black box" trading system you were sold, just create
a phoney stock and price it per the impulse in price.  Run your system
and see what happens.  Repeat for volume impulse.  Repeat for both
price and volume.

One of my clients was using a well-known trading package and complained that
he was getting "yanked around" and couldn't see anything in recent
stock price/volume events that might cause it.

When I did impulse/step tests on this trading package, certain of their
indicators went ballistic about 72 days AFTER the step input.  Wonder why
my client thought he was being "yanked around" by this system?

And my OWN system's development involved step & impulse testing after
every major change in the code.  Just keep equities STEP and PULS around ...

Next chapter, Bob:  Why a simple difference makes a lousy estimator for
rate-of-change.

Cheers,
Robin Lake
Environmental Modeling Inc.
rbl@xxxxxxxxxxx