Page 456 - Design for Six Sigma a Roadmap for Product Development

Fundamentals of Experimental Design 415

response variable is a key performance measure of the process. We
would also want y to be

■ A continuous variable, which would make data analysis much easier
and meaningful
■ A variable that can be easily and accurately measured

Step 3: Choice of factors, levels, and ranges
Actually, steps 2 and 3 can be done simultaneously. It is desirable to
identify all the important factors which may significantly influence the
response variable. Sometimes, the choice of factors is quite obvious, but
in some cases a few very important factors are hidden.
There are two kinds of factors: the continuous factor and the discrete
factor. A continuous factor can be expressed by continuous real num-
bers. For example, weight, speed, and price are continuous factors. A
discrete factor is also known as category variable, or attributes. For
example, type of machines, type of seed, and type of operating system
are discrete factors.
In a DOE project, each experimental factor will be changed at least
once; that is, each factor will have at least two settings. Otherwise,
that factor will not be a variable but rather a fixed factor in the exper-
iment. The numbers of settings of a factor in the experiment are called
levels. For a continuous factor, these levels often correspond to differ-
ent numerical values. For example, two levels of temperature could be
given as 200 and 300 C. For continuous factors, the range of the vari-
able is also important. If the range of variable is too small, then we may
miss lots of useful information. If the range is too large, then the
extreme values might give infeasible experimental runs. For a discrete
variable, the number of levels is often equal to “the number of useful
choices.” For example, if the “type of machine” is the factor, then the
number of levels depends on “How many types are there?” and “Which
types do we want to test in this experiment?”
The choice of number of levels in the experiment also depends on
time and cost considerations. The more levels we have in experimental
factors, the more information we will get from the experiment, but
there will be more experimental runs, leading to higher cost and longer
time to finish the experiment.

Step 4: Select an experimental design
The type of experiment design to be selected will depend on the number
of factors, the number of levels in each factor, and the total number of
experimental runs that we can afford to complete.