Page 151 - Six Sigma Demystified

132 Six SigMa DemystifieD

measured. Manipulating the factors over a wide range provides the best chance
of detecting a change in the process that otherwise may be too subtle to detect.
(A detailed discussion of these techniques is presented in the “Design of Experi-
ments” section in Part 3.)
Consider the case of the villainous cartoon arch nemesis seeking to accu-
rately propel a rock of sufficient size at our hero using yet another fine product
from the ACME Corporation. There are several factors that may be varied to
affect the slingshot’s accuracy, including the size of the rubber band, the draw-
back distance, and the weight of the boulder to be hurdled through the air. The
response variable to be measured is the distance from the large bull’s-eye (sold
separately).
For most designs, and especially in the early stages of the analysis, we can
conveniently limit the design to two levels of each factor. Once we determine
which factors are significant, we can expand the design as necessary, depending
on our objectives.
How does a designed experiment differ from the traditional (sometimes
called one factor at a time) experiments often first taught in grade school? In
traditional experiments, each factor is varied one at a time, such as shown in
Table 6.1. An initial trial (trial 1) is run with each of the factors set at its initial
level. The response (distance from the target) is measured, establishing a
baseline.
A second trial then is run to measure the effect of the first factor (rubber
band size). The difference between the baseline response (trial 1) and the
observed response for trial 2 is assumed to be the effect of the manipulated
factor. In this case, increasing from the small rubber band to the large rubber
band results in a smaller distance from target of 14 units.

Likewise, the effect of the drawback distance is estimated by comparing
trials 3 and 1, and the effect of boulder weight is estimated by comparing trials
4 and 1. In this way, the effect of the drawback distance is estimated as a

TAble­6.1­ Example Traditional Experiment

Trial Rubber Band Drawback Distance Boulder Weight Distance from Target
1 Small Short Light 35
2 Large Short Light 21
3 Small Long Light 28
4 Small Short Heavy 27