This preview shows page 1. Sign up to view the full content.
Unformatted text preview: able Table 2. Response Response variable
(units) Normal operating
level and range Blade profile
(inches) Nominal (target)
+ 1 x 10-3inches to
+ 2 x 10m3 inches at
all points Surface finish Smooth to rough
Typically 0 to 10 Surface defect
count Study. Variables Measurement
precision, accuracyhow known? Relationship of
response variable to
objective u.E= 1 x 1O-5 inches Estimate mean
from target and
Should be as smooth
as possible from a coordinate
standards) TECHNOMETRICS, Must not be
magnitude FEBRUARY 1993, VOL. 35, NO. 1 6 DAVID E. COLEMAN AND DOUGLAS condition (which motivates the experiment). Such a
comparison might be used to derive “performance
measures” from response-variable outcomes. For example, with CNC-machining, blade profile is a response variable, and it is compared to the target
profile by computing differences at certain locations.
Mean absolute difference and the standard deviation
of the differences are performance measures for the
various experimental conditions. They can be analyzed separately or by using the standard deviations
to compute weights for the mean analysis.
5. Is preferably obtained by nondestructive and
nondamaging methods so that repeated measures can
be made and measurement error can be quantified.
6. Should not be near a natural boundary. Otherwise, the variable will not discriminate well. For example, it is hard to distinguish a yield of 99.5% from
99.8%, and it is hard to detect and distinguish contamination levels near 0.
7. Preferably has constant variance over the range
There are other important characteristics of response variables that the experimenters may not have
considered or communicated to the whole experimentation team. This sheet helps to draw them out:
(a) current use, if any (col. 2); (b) ability to measure
View Full Document
- Fall '14
- The Land