This preview shows page 1. Sign up to view the full content.
Unformatted text preview: he analysis.
If a nuisance factor is not measurable and thought
to be very influential, it may also be called an experimental risk factor. Such factors can inflate experimental error, making it more difficult to assess
the significance of control variables. They can also
bias the results. For the CNC-machining example,
the nuisance factors are as shown in Table 5.
Experiment designers have a set of passive strategies (randomization, blocking, analysis of covariante, stratified analysis) to reduce the impact of nuisance factors. These strategies can have a major effect
on the experimental design. They may be constrained Table 4. Held-Constant Factor
(units) 9 EXPERIMENT - Precision of lab
tests unknown How to control
(in experiment) Anticipated
effects Use one type None Do runs after
Use one “midlevel”
Use one lot
(or block on
necessary) None TECHNOMETRICS, None Slight FEBRUARY 1993, VOL. 35, NO. 1 DAVID E. COLEMAN 10 AND DOUGLAS Table 5. Nuisance
known? Nuisance factor
Spindle Vibration of
operation Standard viscosity 1- 2” F. by room
- ? by limits on the number of observations, costs of
changing control-variable settings, and logistic considerations. In the CNC-machining example, the only
nuisance factor to have potentially serious effects and
for which blocking seems appropriate is the machine
spindle effect (though it may be necessary to also
block on titanium forgings). The machine has four
spindles. requiring a design with four blocks or randomizing on all four. Blocking will introduce a bias
in the estimates confounded with the blocking variable(s), whereas randomization will inflate the experimental error. The other two factors are dealt
with by ensuring that they stay below levels at which
problems may be encountered.
7. INTERACTIONS The interactions sheet is self-explanatory. Unfortun...
View Full Document
- Fall '14
- The Land