> ?A>5@ N&bjbj22 *XX.$288888888$RQ8888888Kvvv8.88v8v0v:,8,x:f.,
a06RL88v88888$ vDesign of Experiments (DOE) for Engineers and Technicians
From Levinson Productivity Systems, P.C. www.ct-yankee.com
Audience:
This day-long Design of Experiments workshop is designed primarily for manufacturing engineers, managers, and technicians. There are no statistics prerequisites; the material assumes knowledge of freshman college mathematics only. This overview covers material that participants may encounter in the ASQ certification examination for quality engineering.
The package includes PowerPoint Notes pages for distribution to participants. The license allows the user to make unlimited copies of these pages for use in training activities.
What your participants will learn:
Upon completing this course, your audience will understand:
Why Design of Experiments is useful for improving processes
System models, factors, and levels
The all-important concept of hypothesis testing (which also is the foundation of Statistical Process Control)
How to plan and design an experiment: randomization, blocking, and replication
How to interpret test statistics
Types of experiments and when they are used
How to analyze residuals to see if the experimental model was valid
Linear regression
Nonparametric methods
The course is not a substitute for the kind of in-depth knowledge and problem-solving experience that can only be acquired through college coursework or the equivalent. Since DOE is used far too infrequently in most industrial settings and many (if not most) people don't understand its merits or what the results are telling them, the course will have accomplished its mission if it equips the attendee to at least understand what DOE is, when it is used, and what kind of results it delivers. The material on hypothesis testing should make it far easier for the attendee to succeed in college coursework or self-study of industrial statistics.
Contents
Introduction: what is design of experiments?
Why don't more people use DOE? (A reference explains that many courses have extensive statistical prerequisites, so people don't take them.)
System models, factors, and levels
Determinate (y=f(x)) versus indeterminate (y=f(x)+e) systems.
Hypothesis testing and risks of error in experimentation
A vital concept behind not only DOE but statistical process control (SPC) as well
Planning the experiment
Techniques for excluding extraneous sources of variation
Randomization
Blocking
Replication
Interpreting test statistics
Don't worry about the alphabet soup (including Greek alphabet) of the z, t, c2, and F statistics. They are all applied in pretty much the same way!
Types of Experiments
Single-factor experiments
Tests for the mean of a single population
Two-sided and one-sided tests
Tests for the variance of a single population: c2 test
Comparison of the means of two populations: t test
Comparison of the variances of two populations: F test
Blocking designs
Paired comparison t test
Randomized complete block design
Balanced incomplete block design
Latin squares and Youden squares
Comparison of the means of several populations: Analysis of Variance (ANOVA)
Two-factor experiments
Interactions
Two-way analysis of variance (ANOVA)
Multi-factor experiements
Factorial designs
Fractional factorial designs
Transformations
Testing Residuals
The residual is th9:duw
t
S@$$$$$/%%%L&M&N&ǿǿһh"Oth"OthkUh{Jh"Oth4rhhH*hhOJQJhhlhl6hlhlhqNhl5hqNhlhlh'
h<&:vw
3V4`ST]gd`gdlgd<
&Fgdlgdlgdlgd<N&].tb6(Pz> @Qj
&Fgd4r
&Fgd
&Fgd
&Fgd1>c}$$$$$%/%b%%%%%%L&M&gdk
&Fgd4r
&Fgd{J
&Fgd4r
&Fgd{J
&Fgd{Je difference between the model's fitted value and the actual value. Residuals should follow a normal (bell-curve) distribution.
Chi square test for goodness of fit
Normal probability plot
Linear regression
Practical applications of linear regression
Taylor series assumption
Least squares model fitting
Tests for significance of factors and interactions
Special considerations and pitfalls
Design outliers
Multicollinearity
Autocorrelation
Nonparametric methods
Nonparametric methods often work even when the application does not meet the requirements for traditional methods like ANOVA.
M&N&`gdl/ =!"#$%D@DNormal
`CJ_HmH sH tH L@L Heading 1$$d@&a$5CJ$KHB@B Heading 2$d@&5CJB@B Heading 3$$d@&5H@H Heading 4$$d@&a$5CJDA@DDefault Paragraph FontVi@VTable Normal :V44
la(k@(No List4O4bullet
^CJ>O>Normal_left
$`JOJNumber_table$$H]Ha$CJ6O"6quote]^<O2<Table$]`a$DOBD
Table_list@x^@`:@:
TOC 1h
!5CJ::
TOC 2
!^50Or0
[FootnoteCJ,Oq,
[endnotes6U@67~ Hyperlink>*B*ph.*:vw3V4`ST]:x T b k w (
=
W
@Qj1>c}`
Bfv,-0000000p0000 0 0 0 0 0 0 0 0 0000(0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0( 0 0 0 0 0 0 0 000>c}`
Bfv,-0O900 O900@0O900O900O900O900O900O900O900O900O900O900O900O900O900O900O900O900O900O900O900
0N&]M&N&
N&p|v0:bjl0333V3T]: T (
W
@>c$
`
B0William A. Levinsonh&2"thH.^`o(()
^`hH.
L^`LhH.
^`hH.
XX^X`hH.
(L(^(`LhH.
^`hH.
^`hH.
L^`LhH.h88^8`OJQJo(hHh^`OJQJ^Jo(hHoh ^ `OJQJo(hHh^`OJQJo(hHhxx^x`OJQJ^Jo(hHohHH^H`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHthHh&NҲp
<{J+L
[4r"OtDl7~'
kk@tL
.0@0000,@0H@UnknownGz Times New Roman5Symbol3&z Arial?5 z Courier New;Wingdings"qhVXdB͢<DD%24''
3QH ?'
9Design of Experiments (DOE) for Engineers and TechniciansWilliam A. LevinsonWilliam A. LevinsonOh+'00 LX
t
:Design of Experiments (DOE) for Engineers and TechniciansnesiWilliam A. LevinsonillillNormal.dotLWilliam A. Levinson12lMicrosoft Word 10.0@ha@0!@x:D՜.+,0 hp|
n'A
:Design of Experiments (DOE) for Engineers and TechniciansTitle
!"#$%&'()*+,-/012345789:;<=@Root Entry Fx:BData
1TableWordDocument*SummaryInformation(.DocumentSummaryInformation86CompObjj
FMicrosoft Word Document
MSWordDocWord.Document.89q