Page 244 - Six Sigma Demystified

224 Six SigMa DemystifieD

The best, and most common, approach is to begin with an effective screening
design, which will provide information on key factors and the two-factor inter-
actions between those factors. The design can be improved sequentially with
additional trials to acquire additional information.

Defining Responses
The response in a designed experiment is the parameter you are observing as the
outcome of the experiment. For example, in a manufacturing process, you may
be concerned about the density of an injection-molded part. You will change a
variety of conditions within the process and measure the resulting part density.
In a service process, you may seek to measure the impact of process changes on
customer satisfaction or cycle time.
You often can measure more than one response in an experiment. In some
cases, the responses are converted or transformed during analysis to simplify the
model or to uncover factors that contribute to variation in the response. See
also “Transformation” below.
In any event, the most useful responses are quantitative variables rather than
qualitative attributes. You need sufficient resolution on the measured variable
to use the statistical regression techniques. Before conducting the experiment,
you should analyze the measurement system for error in estimating the response
using measurement systems analysis (MSA) techniques. When responses are
qualitative, you sometimes can convert them to quantitative scores (such as
Likert scales). When conversion is not convenient or helpful, logistic regression
techniques should be used for analysis.

Defining Factors

The parameters that you vary in the process to achieve changes in the response
are known as factors. Generally, you will control these factors by setting them
at specific levels for each run, or trial, of the experiment. You will run the ex-
periment at various conditions of each of the factors so that the effect of each
factor on the response can be calculated. In an injection-molding process, you
may want to investigate the effect of changing furnace temperature, fill pres-
sure, and the moisture of the raw materials. Even though you generally cannot
set the moisture level of the materials in normal operations, you can sample and
segregate the materials into two or more distinct levels for the experiment.
Likewise in a service process, you may choose, for the experiment, to test the
effect of two different process designs, such as with and without customer
follow-up.