A Scientific Approach
to Determining Root Cause
Bev Daniels
The core principle of Six Sigma is Problem Solving of chronic problems, yet this is the
weakest area of the initiative. Analyze phase is best described as ‘then a miracle
occurs’
This presentation is intended to provide an introduction to a disciplined structured
approach to determining root cause. This method is based on sound scientific and
statistical principles.
Although we will only be able to cover enough material to provide an overview, there is
substantial extra materials and references/bibliography for further study.
What are some of the problem solving methods – not statistical tools – that you have
used?
Slide 2
ACT PLAN
PDSA
STUDY DO
8D - 8 Disciplines DMAIC
•Plan
•Define
•Form a Team
•Define the Problem •Measure
•Containment •Analyze
•Identify Root and Escape Cause •Improve
•Choose Permanent Solution •Control
•Implement & Validate Solution
•Preventive Measure for Systemic Causes
•Congratulate & Celebrate
This presentation focuses only on the “Analyze” phase. Also known as known as the
“Identify Root Cause” or Root Cause Analysis step of Problem Solving.
It does not address how to develop and validate solutions, implement controls or
measurement systems analysis.
Slide 3
Physics
or
Geometry
Problem
Business
(People)
Processes
Problems can be broken down into two basic types: Physics based technical problems
or People based process problems. This distinction applies to the Problem or Effect and
not to the cause side of the causal system. These are operational definitions and as
such are somewhat loose and open to interpretation. But in general they will serve us
well in helping to determine the appropriate Problem Solving Strategies.
Theory based
• Brainstorming & fishbone diagrams
• Scientific theories of how a specific factor or event
creates the problem (e.g. fault tree)
FMEA
Trial and error – often solution based
Experience based
5 Why
“Is, Is Not” (Kepner-Tregoe)
Brainstorming often involves multi-voting and other consensus based ranking schemes
to select theories to test. “science is not a democracy” “public opinion polls have never
changed a law of physics”
FMEA has elements of Problem Solving but is not intended to be a diagnostic tool.
FMEA is focused on how multiple functions fail (complete, partial, intermittent or
unexpected) and what might cause each one as well as severity, occurrence and
detection ratings. In a very limited sense the concept function failure is useful in
defining the Problem and the detection rating is a result of a MSA. The listing of
causes is still limited to known causes or guesses.
Cause to Effect: The literature has a lot of references to the various techniques, but
little rigorous explanation of how to use the techniques.
Effect to Cause: The literature has few references to the various techniques but tend to
be much more detailed and rigorous in their use. However, the references are very
disparate and there is very little in the literature that synthesizes the techniques in a
coherent and systematically useful manner.
Slide 6
Cause to effect can work, it’s just not very efficient (low batting average).
It’s very seductive since it often promises a quick discovery of the cause or a solution.
What most people forget is that there are too often many iterations of the ‘quick’
approach and too often the solution doesn’t work.
Problem
Apollo
“Sequence of Events” Business
(People)
Processes
Lean & 5-Whys
Y→X is a 5-Why strategy that utilizes the principles of analytical studies (Deming) and
the “half-split”* technique to work backwards in an iterative progression from the
Problem statement to converge on the causal mechanism.
Immediate
Cause
The Ys CTQs Metrics
Immediate
Cause Structural Y Functional Y Customer Y $$
Although the individual factors are highly confounded in the begining, it is easy to
unconfound the single category that contains the causal mechanism.
Definitions
Failure modes
True defects
Stress Strength
Distribution of the
Distribution of the
Primary X for the
severity of the
product
conditions that the
product can see Parts that fail
Increasing “strength”
Strength
Old New
Stress
No parts fail!
Once the conditions for failure are known and defined all
subsequent testing must take place under the same
conditions that can cause failure
“Root cause”
In people based processes the “root cause” is often a combination of a specific action
and a specific condition, that may result in a cascading series of actions and conditions.
(See the Apollo method by Dean Gano)
Slide 19
Other causes
Peripheral causes
Until the causal mechanism is determined and understood, discussion of these causes
are diversionary; addressing them will not solve the current Problem
Enabling causes can and should be addressed during the diagnostic process IF they
can improve the measurability of the Problem or enable better screening to provide
effective containment of the Problem.
Systemic causes cannot typically be effectively addressed until the causal mechanism
is known. Discussion of a systemic casue for an unknown physical cause is conjecture.
Slide 21
Sequence of causes
Enabling Causes
Condition
for Failure
Time Sequence
Enabling Causes
Condition
for Failure
Time Sequence
The
Why did the Titanic
Titanic sink? Sinks
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 23 IDEXX Laboratories, Inc.
Slide 24
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 24 IDEXX Laboratories, Inc.
Don’t skip the ‘obvious’; if it’s a incontrovertible fact, document it; if it’s an assumption,
test to disprove.
Slide 25
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 25 IDEXX Laboratories, Inc.
An enabling cause – the smaller bulkheads filled with water faster than larger bulkheads
Slide 26
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 26 IDEXX Laboratories, Inc.
Swamped
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 27 IDEXX Laboratories, Inc.
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 28 IDEXX Laboratories, Inc.
The Titanic’s sister ship had a side swipe accident in shallow waters and her seems
opened up. The thumbnail above is a picture of this failure.
Unfortunately, the side of the ship that was damaged by hitting the iceberg is the one on
the ocean floor and so we have no direct evidence that it too opened up. We do
however, have additional evidence that makes this factor the most likely cause of water
entering the ship.
Slide 29
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 29 IDEXX Laboratories, Inc.
Rivets found in the debris field were found in great numbers to have been sheared at
the rivet head. Shear forces are very easy to identify under metallurgical examination.
Eye witness – ear witness – accounts reported hearing ‘popping’ noises like ball
bearings or marbles hitting the floor at the time of the impact.
Sheared rivets are almost conclusive proof that the seams opened up rather than the
metal sheets being ‘torn’ open
Slide 30
Metallurgical structure
Open Seams on Olympia
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 30 IDEXX Laboratories, Inc.
Metallurgical examination revealed the presence of extensive slag in the rivets which is
known to weaken the strength of metal.
But the rivets didn’t shear on their own
Slide 31
Metallurgical structure
Open Seams on Olympia
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 31 IDEXX Laboratories, Inc.
Metallurgical structure
Open Seams on Olympia
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 32 IDEXX Laboratories, Inc.
Now we are dealing with the specific event. We are also crossing over from physics to
people based systems.
All physics based failures have a systemic cause in people based systems.
People create many of the specific conditions for failure* and people certainly create the
conditions for behavioral causes (actions).
*People didn’t create the atmospheric condition that made the iceberg practically
invisible until the last seconds, nature did. But people did create an environment that
led to reckless behavior in entering the ice field and trying to run as fast as possible…
Slide 33
Metallurgical structure
Open Seams on Olympia
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 33 IDEXX Laboratories, Inc.
Slide 34
Metallurgical structure
Open Seams on Olympia
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 34 IDEXX Laboratories, Inc.
Slide 35
Metallurgical structure
Open Seams on Olympia
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 35 IDEXX Laboratories, Inc.
Slide 36
Metallurgical structure
Open Seams on Olympia
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 36 IDEXX Laboratories, Inc.
Slide 37
EFFECT
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 37 IDEXX Laboratories, Inc.
This is a relatively simple example that really didn’t require sophisticated experimental
designs to answer any of the 5 Why questions. It does show the basic flow of causal
mechanism and the separation of systemic, enabling and physical causes.
There are a great deal of other factors that contributed to the Titanic being in the
presence of this massive iceberg during a thermal inversion. There are also factors that
made the death toll higher than it should have been.
It is a romantic notion to say that “no one thing caused the Titanic to sink” or that “there
were a cascade of unforeseeable events that doomed the Titanic to its tragic demise”.
Certainly these events cascaded to put the Titanic on a collision course with the iceberg.
But to give this series of events such importance is to give up control of our lives to the
fates. Nothing could be further from the truth. The Titanic sank because it experienced
a shear force and it had very weak rivets that could not withstand this force. If it hadn’t
hit the iceberg on April 14 1912, it would have hit something at some time and the
seams would have opened up then. Increasing the strength of the fastening system
would have saved the Titanic or future ships of similar size and this is the goal of
engineers and scientists.
Slide 38
Immediate
Cause
Immediate
Cause
Immediate
Cause Structural Y Functional Y
Structure Function
• Dimensions • Performance
• Properties tolerances
Immediate
• Esthetics
Cause
Immediate
Cause
Branch of Causes
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 38 IDEXX Laboratories, Inc.
Slide 39
Disprove
Immediate
Cause
Immediate
Cause
Immediate
Cause The Y Functional Y
Immediate
Cause
Factor
Immediate
Factor Factor
Cause
Factor 2nd Cause
Root Cause
8 9 10 11 12 13 14 15 16 17
18 19
20 21 22 23 24 25 26
27 28 29 30 31 32
33 34 35 36 37 38 39 40
41 42 43
44 45 46 47 48 49 50 51 52 53 54
55 56
57 58 59 60 61 62 63
64 65 66 67 68 69 70 71 72
73 74 75 76 77 78 79
80 81
82 83 84 85 86 87 88 89 90 91 92
93 94 95
117 118
119 120 121 122 123 124 125 126 127 128
129
This example is a simple linear flow with no interactions and only one causal factor.
It also represents a deterministic approach (the failure rate is a binary response). If this
were a real world example, the failure rate will most likely be a consideration.
If there were a secondary cause (less effect than a primary cause, the investigator
would detect this as the failure rate would not go to zero when the primary cause was
determined and/or controlled.)
This causal system is amenable to the half split technique.
Other systems may require multi-way splits.
Slide 43
Split categories
Some causal systems will use hybrids of the 3 primary split categories
Slide 44
Location
• Within piece
• cavity to cavity
• station to station
• line to line
• plant to plant
• region to region
Components
• Sub-assemblies
• Components
• Raw materials
• Process (assembly or process methods)
Specific features, dimensions and/or properties
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 45 IDEXX Laboratories, Inc.
Slide 46
Temporal
• (Within Piece), Piece to piece, lot to lot, vendor lot
to lot, month to month, season to season etc.
• Product use: during use, use to use
• Operator to Operator
• Within a process; step to step or operation to
operation.
Diagnostic pairs5
1. What’s Wrong
What’s
Happening?
2. What’s Changed
Effectiveness
What’s
Different?
3. What’s Different
What’s
What’s
Wrong?
What’s Wrong: The most common approach and works well with simple Problems and
obvious (easily observed) causes.
What’s Changed: The weakest question. It requires that the cause was known,
measured and recorded. It is also prone to post hoc, ergo propter hoc* errors.
What’s Different: Involves determining the differences between the diagnostic pairs.
When used with a convergent elimination strategy it is highly effective.
What’s Happening: This is the strongest question. It includes the other three
questions as appropriate and when coupled with a convergent elimination strategy is
the most effective approach for highly complex problems
Slide 51
It is always helpful for the team lead – and team members where feasible – to carry
around the part experiencing the Problem. Holding it, looking at it, playing with it, keep
the eladers focus on the Problem part and it’s function.
Slide 56
Types of Studies
Enumerative Analytic
Descriptive Predictive
• Estimation of a finite population or static
data set • Understand the causal mechanisms and
resulting performance of a system in
• Quantifies only the product or process in
order to make predictions about future
front of us.
performance
• Has no predictive usefulness for future
performance. • Used to improve products and processes
in the future
Statistics and statistical precision of the Tests of statistical significance are often
estimates have value. redundant
Used to determine what action should Proper structure is more important to
be taken on the population under study. our belief in the prediction than
statistical estimates of the precision or
accuracy of the analysis.
Examples: Census, customer surveys, Examples: Y-X problem solving studies,
acceptance sampling, factorial experiments, response surface
experiments
Critical Structural Element: Critical Structural Elements:
Randomness (representativeness) of Independence & Replication
the sample
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 56 IDEXX Laboratories, Inc.
Analytic Studies
Replication
Science &
Engineering
6
Practical
Statistics
Knowledge
A classic example is The Dictionary Game: One player picks a word from the dictionary
and the other player tries to guess the word. The guessing player then attempts to
determine what the selected word is by asking questions that can only be answered
with a ‘yes’ or a ‘no’. The weakest approach is to ask very specific questions: “is it this
word or that word”. They may elect to ask such questions as “does it start with an ‘A’”
or “is it an animal, vegetable or mineral”, or perhaps even “is it a noun, verb, descriptor
or interjection”? While each of these questioning strategies has some ‘elimination’
power, they are weaker than simply asking “is the word in the first half of the dictionary”
and progressively cutting the part of the dictionary that contains the word in half until the
guessing player is down to the last remaining word. The player doesn’t need to know
what the word is, how to pronounce it or even how to spell it; they simply need to have
the dictionary in their hands.
Slide 62
Balance
Structure
Sample size and independent replicates
Simultaneous testing of all alternative theories
• Use of appropriate experimental controls: ensure
that external extraneous changes are not missed or
misinterpreted.
• Inclusion of changes in “experimentally uncontrolled”
factors
• Randomization
Protocol:
• ensuring that the test conditions replicate
normal operating conditions
• Ensuring that worst case conditions are
correlated to normal conditions and are not
beyond actual worst case conditions
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 64 IDEXX Laboratories, Inc.
Slide 65
Discipline
A turned dimension
The Problem:
What are the ‘categories’
A 3 Component assembly is
experiencing cracks in the clips
for the first level?
and some loose B Parts...
Part A
Top View
Part B
The Y
Cracked
Clip Clips
Clip
Side
View A B
A is too big
The Y
B is too big
Cracked
Clips
C is too small
C is too weak
First experiment
Part A
Top View
field.
A “post hoc” analysis was performed Clip
The Results
Cracked Cracked
Center Center
0 1 2 3 4 5 6 7 8 9 10 11 0 1 2 3 4 5 6 7 8 9 10 11
Assembly Assembly
Cracked Cracked
Center ` Center
0 1 2 3 4 5 6 7 8 9 10 11 0 1 2 3 4 5 6 7 8 9 10 11
Assembly Assembly
This comparison can also be displayed on a Youden plot. A systemic difference will
display as a bias to a 45 degree 1:1 line
Slide 72
Side
View A B
Tk
Clip
Within Piece
Piece - Piece = Within Fixture
Set to Set
Machine to Machine
Operator to Operator
Time to Time
Shift to Shift
Vendor Batch to Batch
MSA
Cracked
Center
Within Piece
0 1 2 3 4 5 6 7 8 9 10 11
Assembly
Clip A B
Lot to Lot
The
B is too big Y
Time to Time
Cracked
Operator to Clips
Operator C is too small
Lathe to Lathe
C is too weak
Vendor Lot to
Lot
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 75 IDEXX Laboratories, Inc.
Slide 76
Multi-Vari results
Data was taken from 3 sets, 3 readings per part (at each of the 3
clip locations), on all 8 parts off of a single machine. Two
separate operators were involved. The results were:
10 10
9 9
8 8
7 7 Left
6 6 Cent er
5 5 Right
4 4 LSL
3 3 USL
2 2
1 1
0 0
3
12
15
18
21
24
27
Which category causes the largest change in the Y?
Do you see anything unusual about the data?
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 76 IDEXX Laboratories, Inc.
With a multi-vari when the majority of the baseline variation in Y is observed, root cause
was active in that category and it is not necessary to proceed to test the other
categories
Slide 77
MSA
Cracked
Center
0 1 2 3 4 5 6 7 8 9 10 11
Clip A B
Lot to Lot
The Y
B is too big
Time to Time
Cracked
The centers are thin Operator to Clips
Operator C is too small
The ends are thick
How does this happen? Lathe to Lathe
10
9
8
7
10
9
8
7 Left
C is too weak
6 6 Center
5 5 Right
4 4 LSL
3 3 USL
2 2
1 1
0 0
3
12
15
18
21
24
27
Clip A B
Lot to Lot
The Y
B is too big
Time to Time
Cracked
Operator to
Clips
Operator C is too small
Lathe to Lathe
10
9
8
7
10
9
8
7 Left
C is too weak
6 6 Center
5 5 Right
4 4 LSL
3 3 USL
2 2
1 1
0 0
3
12
15
18
21
24
27
10 10
9 9 Left
8 8 Center
7 7
Right
6 6
New Left
5 5
New Center
4 4
3 3 New Right
2 2 LSL
1 1 USL
0 0
3
12
15
18
21
24
27
33
36
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 79 IDEXX Laboratories, Inc.
Slide 80
MSA
Clamping Cracked
Strength 0 1 2 3 4 5 6 7 8 9 10 11
Cutter moving
10 10
Piece to Piece A is too big Clip
9 9 Left
8 8
7 7
Center Clip A B
Right
6 6
New Left
5 5
New Center
4 4
Lot to Lot
3 3 New Right
2 2 LSL
The
1 1 USL
0 0
12
15
18
21
24
27
33
36
B is too big Y
Time to Time
Cracked
Operator to Clips
Operator C is too small
Lathe to Lathe
10
9
8
7
10
9
8
7 Left
C is too weak
6 6 Center
5 5 Right
4 4 LSL
3 3 USL
2 2
1 1
0 0
3
12
15
18
21
24
27
Notice the thumbnails of the critical analyses…although they are not easily read on the
diagram, they serve as ‘objective evidence’ that each level is closed with data, not
opinion.
Slide 81
Second Example
A prototype printer
A new low cost printer design. Each of 5 prototypes are exhibiting ~ 5%
misfeeds (multiple sheets pulled at a time resulting in a paper jam:
What analysis strategy & questions would you use to get to Root Cause?
Direction of Paper Travel
Paper Roller
Paper Stack
Retard Pad
Within Stack
Job Type:
Large or Small
Frequency:
Short or Long Time Paper Jams
Between Jobs
Stack to Stack
Paper Type to
Type
Printer to Printer
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 85 IDEXX Laboratories, Inc.
Slide 86
Page to Page
Within Stack
Job Type:
Large or Small
Frequency:
Short or Long
Paper Jams
What changes within the Time Between
Jobs
stack?
Direction of Paper Travel Stack to Stack
Paper Roller
Paper Stack
Paper Lot to Lot
Retard Pad
Tray Lift Spring
Paper Tray
Paper Type to
Type
5% Misfeeds
ASQ Lean Six Sigma Conference Printer to Printer Bev Daniels
February, 2011 88 IDEXX Laboratories, Inc.
Slide 89
Retard Pad
Tray Lift Spring
Paper Tray
Retard Pad
Tray Lift Spring
Paper Tray
5% Misfeeds
Slide 91
Frequency:
Short or Long
Paper Jams
Time Between
Jobs
Stack to Stack
Paper Type to
Type
Retard Pad
Retard Pad
Tray Lift Spring
Paper Tray
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 92 IDEXX Laboratories, Inc.
5% Misfeeds
Slide 93
Frequency:
Short or Long
Paper Jams
Time Between
Jobs
Stack to Stack
Paper Type to
Type
‘excursionary’ change in a 17
16
15
14
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
that create a stable process
that behaves randomly within
predictable limits.
An Assignable Cause isn’t necessarily easy to find and correct
Exercise: cards10
The Rules:
• Each card has a letter on one side and a number on the
other
• If the card has a letter that is a vowel then the number
will be an even number
The Game: Given the following 4 cards which cards would
you flip to determine if the rule were true or not? You may
only flip 1 or 2 cards:
A 7 8 P
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 98 IDEXX Laboratories, Inc.
Slide 99
A 7 8 P
References
[1] Steiner, Stefan H., MacKay, R. Jock, “Strategies for Variability Reduction”, Quality Engineering, Volume
10, Issue 1, September 1997 , pp 125-136
[2] Kavuri, Surya N., Rengaswamy, Raghunathan, Venkatasubramanian, Venkat, “A Review of Process Fault
Detection and Diagnosis Part II: Qualitative Models and Search Strategies”, Computers and Chemical
Engineering, 27 (2003), pp. 313-326
[3] Smith, Gerald, F., “Determining the Cause of Quality Problems: Lessons From Diagnostic Disciplines”,
Quality Management Journal, 98 5, No. 2, 1998 ASQ
[4] Dale, H. C. A., “Fault Finding in Electronic Equipment”, Ergonomics, pp. 356-383, 1957
[5] Charles Higgins Kepner, Benjamin B. Tregoe, The New Rational Manager, Princeton Research Press,
1981
[6] Allen, John R., “Three Good Questions (and One Not So Good), The New Science of Fixing Things, 2006,
www.tnsft.com
[7] Youden, William John, “Graphical Diagnosis of Interlaboratory Test Results”, Industrial Quality Control,
May 1959, Vol. 15, No. 11
[8] Donald S. Ermer and Robin Yang E-Hok, “Reliable data is an Important Commodity”, The Standard, ASQ
Measurement Society Newsletter, Winter 1997, pp. 15-30.
[9] Donald J Wheeler, “An Honest Gauge R&R Study”, Manuscript 189, January 2009.
http://www.spcpress.com/pdf/DJW189.pdf
[10] Kida, Thomas, Don’t Believe Everything You Think: The 6 Basic Mistakes We Make in Thinking,
Promethius Books, 2006
Gano, Dean, L., “Effective Problem Solving – A New Way Of Thinking”, Apollo Associated Services, Inc.,
www.apollo-as.com
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 100 IDEXX Laboratories, Inc.
Slide 101
Bibliography
Seder, Leonard, “The Technique of Experimenting in the Factory”, Industrial Quality Control,
March 1948
Gano, Dean, Apollo Root Cause Analysis - A New Way Of Thinking, Apollonian Publications,
Distributed by BookMasters, Inc., 1999
Allen, John R., Hartshorne, David J., “The Art and Science of Fixing Things”, 2006,
www.tnsft.com
Steiner, Stefan H., MacKay, R. Jock, Statistical Engineering: An Algorithm for Reducing
Variation in Manufacturing Processes, ASQ Quality Press, 2005
Haviland, Paul R., “Analytical Problem Solving”, The Haviland Consulting Group,
www.fmeca.com
Leonard A. Seder, “Diagnosis With Diagrams – Part I”, Industrial Quality Control, January 1950,
pp. 11-19
Leonard A. Seder, “Diagnosis With Diagrams – Part II”, Industrial Quality Control, March 1950,
pp. 7-11
Mario Perez-Wilson, Multi Vari – A Pre-Experimentation Technique, Advanced Systems
Consultants, 1992
Robert D Zaciewski and Lou Nemeth, “The Multi-Vari Chart: An Underutilized Quality Tool”,
Quality Progress, October 1995, pp. 81-83
Types of Studies
Enumerative Analytic
Descriptive Predictive
• Estimation of a finite population or static
data set • Understand the causal mechanisms and
resulting performance of a system in
• Quantifies only the product or process in
order to make predictions about future
front of us.
performance
• Has no predictive usefulness for future
performance. • Used to improve products and processes
in the future
Statistics and statistical precision of the Tests of statistical significance are often
estimates have value. redundant
Used to determine what action should Proper structure is more important to
be taken on the population under study. our belief in the prediction than
statistical estimates of the precision or
accuracy of the analysis.
Examples: Census, customer surveys, Examples: Y-X problem solving studies,
acceptance sampling, factorial experiments, response surface
experiments
Critical Structural Element: Critical Structural Elements:
Randomness (representativeness) of Independence & Replication
the sample
ASQ Lean Six Sigma Conference Bev Daniels
February, 2011 102 IDEXX Laboratories, Inc.
Slide 103
Enumerative Studies
Analytic Studies
Replication
Types of Factors
• Direct Control
• Indirect Control
• Uncontrolled
Direct Control
Indirect Control
Indirect Control factors are also referred to as “chunky type” or “background” factors.
This type of grouping of factors is also used in Components of Variation studies.
Slide 111
Indirect Control
Indirect Control factors are also referred to as “chunky type” or “background” factors.
This type of grouping of factors is also used in Components of Variation studies.
Slide 112
Indirect Control
Indirect Control factors are also referred to as “chunky type” or “background” factors.
This type of grouping of factors is also used in Components of Variation studies.
Slide 113
Uncontrolled
Unknown factors
Customer usage in this context involves selctign use conditions by selecting different
customers as in a field trial or Beta testing.
Use conditions, once identified can be directly controlled either as a direct control main
effect or in a grouping or ‘block’ by selecting several use conditions and controlling
each condition to a unique level.
Slide 115
Applied energy is not uniformly distributed across the target area (among cavities, within
a chamber, across a web) and/or over time.
Mixtures are not homogenously distributed within the volume of material
Slide 116
Unknown
All Factors
1 Deming, W. Edwards, “On Probability as a Basis for Action”, The American Statistician, 1975,
Vol. 29, No. 4, pp146-152
2 Moen, Ronald D., Nolan, Thomas, W., Provost, Lloyd P., “Quality Improvement through Planned
Experimentation”, 2nd Edition, McGraw-Hill, 1999
3 Kida, Thomas, “Don’t Believe Everything You Think, Prometheus Books, 2006
4 Bauernfeind, R. H. “The Need for replication in Educational Research”, Phi Delta Kappan, No.
50, pp. 126-128, 1968
5 Sterne, Jonathon A. G., Smith, George Davey," Sifting the Evidence – What’s Wrong With
Significance Tests?”, British Medical Journal, Volume 322, January 2001, pp226-231
6 Johnson, Douglas H., Journal of Wildlife Management, “The Insignificance of Statistical
Significance Testing”, Vol. 63, Issue 3, pp. 763-772, 1999
7 Carver, Ronald P., “The Case Against Statistical Significance Testing”, Harvard Educational
Review, Vol 48, Issue 3, pp 378-399, 1978
8 Shaver, James P., “What Statistical Significance Testing Is and What It Is Not”, Journal of
Experimental Education, No.61, pp. 293-316, 1993
9 Tukey, John W., “A Quick, Compact, Two Sample Test to Duckworth’s Specifications”,
Technometrics, vol. 1, February 1959, pp. 31-48
10 Gans, D. J. (1981), “Corrected and Extended Tables for Tukey’s Quick Test”, Technometrics, 23,
pp 193-195
11 Westlake, W.J. (1971), “A One-Sided Version of the Tukey-Duckworth Test”, Technometrics, 13,
p 901-903