Quality Money Management : Process Engineering and Best Practices for Systematic Trading and Investment

267

29.6. Analysis of Variance

ANOVA is a collection of statistical models where observed variation is partitioned into

components explained by different inputs plus an error term. What is called one-way

ANOVA compares several independent groups of observations, with potentially differing

means and variances. A test can prove whether all the means are equal. What is called

two-way ANOVA studies the effects of two factors separately, including their individual

main effects and their interaction effect. Total variation, SST , can be partitioned into a

component, SSR , that represents variability which is explained by the slope of the regres-

sion equation, and another component, SSE , that represents the random or unexplained

deviation of points from the regression line. That is:

SST SSR SSE yi y

i

n





()^2

1

∑

where

SSR yi y

i

n





()ˆ^2

1

∑

and

SSE yiiy

i

n





()ˆ^2

1

∑

Design of experiments techniques provide an efficient means for product teams to opti-

mize a mixture process. The organization must keep records of nonconforming trading

systems, including the types and nature of nonconformities, and actions taken to fix the

nonconformities, or to preclude the system ’ s continued use. Before it can be put to use,

the organization must reverify the implementation of the containment of fix by stepping

through the appropriate stages of K|V.

29.7. Sources of Assignable Causes

The root causes of special, or assignable, variation in trading/investment systems often

come from one of the following (in no particular order):

● Legal changes, such as Sarbanes-Oxley.

● Regulatory or GAAP changes.

● New, competing systems come online.

● Changes in exchange matching algorithms.

● Technology changes, such as new server technology or FIX or FAST version changes.

(A trading/investment machine, unlike a real, physical machine, does not wear down;

parts do not become corroded. Technology can, however, become obsolete.)

● Economic shifts.

29.7. SOURCES OF ASSIGNABLE CAUSES