15 black belt week three student copy - lean ohio · black belt transforming the public sector week...
TRANSCRIPT
9/16/2015
1
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.govSIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Black BeltTransforming the Public Sector
Week Three
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Code of Conduct• If you have a question – ask• If you have an experience or real data that
relates to the topic - please share it• We have frequent short breaks - be prompt in
returning• Please be courteous to your neighbors silence
your cell phones• Please, no sidebar conversations. Share your
thoughts• There will be plenty of team activities - please
participate
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Code of Conduct
• Please listen – think how can I use this tool or knowledge?
• Record all pertinent ideas on “Aha sheet”• Please limit e-mail or accessing the network
during training • It is okay to disagree• Attack issues & problems – not people• Have fun!
9/16/2015
2
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Ground Rules
• Everyone participates• Open and honest dialogue
• Respect Opinions• Consensus
• Leave Rank at the Door• No Silent Disagreement• Blameless Environment
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Support Materials
• Laptop• Software:
– Minitab– Microsoft Office/Excel– Visio
• Presentation files• Examples & handouts
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Icebreaker
9/16/2015
3
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Icebreaker Instructions
• On a slip of paper please write something unique about yourself (a good story to share)
• Pick a story from the box and then find the person in which the story belongs
• Be prepared to report out introducing that person (name and agency) and then tell that persons story to the rest of the class
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Six Sigma Review
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Lean Six Sigma Methodologies
• Used for improving process efficiency & reducing cycle time.
LeanProcess Efficiency & Speed
• Used for getting quick process improvement, 5 days, good for recurring issues.
Kaizen EventsQuick Process Improvements
• Used for improving quality, reducing variation, and eliminating defects.
DMAICVariation & Defect Reduction
• Used on smaller sub-processes - routine processes.Lean RoutineSub-processes
• Used for developing new products or processes; or to radically change in process.
3PProduction Preparation
Process
• Used in strategic planning to map the entire value stream all product families.
Value Stream Mapping
9/16/2015
4
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
DMAIC Tools
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Flow of Week
• Monday: Define• Tuesday: Measure:
• Wednesday: Analyze• Thursday: Improve/Control
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Six Sigma Overview
Six Sigma is a set of tools and strategies for process improvement that seeks to improve the quality of process outputs by:• identifying and removing the causes of defects
(errors) • and minimizing variation in processes
9/16/2015
5
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Y = f(X1, X2, X3…)
The Framework for Six Sigma
The Six Sigma Approach
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.govCritical Input Variables
30+ Inputs
8 - 10
4 - 8
3 - 6
Found Critical X’s
Controlling Critical X’s
10 - 15
All X’s
1st “Hit List”
Screened List
MEASURE
ANALYZE
IMPROVE
CONTROL
• Multi-Vari Studies
• Design of Experiments (DOE)
• Control Plans
• Process Maps• C&E Matrix• FMEA• Decision Tree• Process Capability
The Funneling Effect: Out to Lunch
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Six Sigma as a Philosophy• Focused on what matters to the customer
• The basic premise of Six Sigma is that sources of variation can be:– Identified– Quantified– Eliminated or controlled
• Focused on strategic or core processes
• Data driven– Measurements focused on the right things
9/16/2015
6
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Six Sigma as a Strategy
Gain insight into our current operational processes
� How good is it (Baseline)?
� How good could it be (Entitlement)?
� What is limiting it from getting better?
� How can it be improved (Closing the Gap)?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
2
3
456
308,537
66,807
6,210
233
3.4
σ DPMO
Six Sigma as a Concept• “Sigma” measures the ability to
meet performance requirements
• Six Sigma methods can be applied to all business functions– Products and Operations– “Technical” issues– Service and Transactional– “Administrative” issues
• Six Sigma represents a performance goal and implementing problem-solving methods
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Reducing variability is the essence of Six Sigma
Every Organization Activity Has Variability...
Mean
1s
Target
(defect)
UpperCustomer
Specification
LowerCustomer
Specification
Variation is Evil
9/16/2015
7
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov19
1s
Much Less Chance of
Failure
Reducing Variation is Reducing Variation is the key to Reducing
Defects
6s
m
1s
Some Chance of
Failure
3s
MeanMean Specification Specification Limit
Variation is Evil
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
LSL
Short-Term
Capability
Long-Term Capability
Short-Term
Capability
USL
Shift Happens
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
StructureReporting
Requirements
9/16/2015
8
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Leadership Lean Six Sigma Roles
Project Roles
• Sponsor(s)
• Lean Liaison
• Process Owner(s)
• Subject Matter Experts (SMEs)
• Team Members
• Fresh Perspective
• Customer
• Corgwn
Belts
• Master Black Belt
• Black Belt
• Green Belt
• Camo Belt (Lean)
• Yellow Belt
• White Belt
• Mentor
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Key Requirements of Black Belts
• Technical Six Sigma expert • Deliver results on schedule• Document learnings• Attend all 5 weeks of
training• Show relevant data• Identify barriers to
progress• Lead team to execute
projects
• Stimulate Leadership thinking
• Determine appropriate tools to apply
• Prepare a detailed project assessment during measurement phase
• Identify project resources (project management)
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
• Report progress to leadership
• Solicit help from Sponsors/Champions
• Influence without direct authority
• Present final results• Insure results are
sustained • Manage project risk
• Get input from SMEs, front line workers, teams and coaches
• Lead way in implementing Lean
• Teach and coach Lean Six Sigma methods and tools
Key Requirements of Black Belts
9/16/2015
9
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
• Sponsor & Mentor Review– Timely, consistent project reviews are a requirement for
successful projects
• Quarterly LeanOhio Project Report-out
Outline of Reporting Requirements
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
DEFINE
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Define
• Define Purpose: To identify and prioritize the improvement opportunity, define critical customer requirement, document the processes and build effective teams.
1. Identify Critical to Quality2. Develop Project Charter
3. Visualize the Process
9/16/2015
10
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Define Tools
• Project Charter• Team Identification
• 4 Voices• SIPOC
• Identify Measures/ Baseline Data
• Team Formation
• Scoping Meetings• Project Benefits
• CT Flow Down • VOC – CTQ
• Project Diagnostics • Team Management
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Help Desk Case Study
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Simulation Introduction
• Department of Prevention• Processing IT Help Desk Process
• Data, additional information and questions are provided for each step of the process and will follow this week of training
*Please do not read ahead in the simulation
9/16/2015
11
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Report Out: Thursday
• Throughout the week save output of activities and/or pictures in powerpoint for report out on Thursday
• Simulate Black Belt Report out using template provided
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Black Belt Report Out Template
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
• Start creating your report out now• Conduct report out the simulation tools
used this week • Add everything and then delete or edit
what you don’t need on Thursday!• Assign one person/laptop to keep
materials• Have everyone in your team conduct
minitab/excel calculations
Report Out Tips
9/16/2015
12
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Scoping MeetingsDMAIC
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Pre-Scoping Meeting(s)
Pre-Scoping• Process Sponsor• Process Owner• FacilitatorsOutput• SIPOC • Possibly Identification of Project Type• Assignment of Project Charter• Initial Call for Baseline Data
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Scoping Meeting(s)
Scoping• Process Sponsor• Process Owner• Facilitators• Other team members if neededOutput• Identification of Type of Project• Completed initial Project Charter (Scoping Document)• Identification of Team Members• Dates/Logistics • Baseline Data
9/16/2015
13
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Review Scoping Timeline
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Help Desk Background/ Initial Scoping
Meeting Questions
• Review the *Case Study Information*: Help Desk Background Information
• In your teams complete the Initial Scoping Meeting Questions
• Obey the Stop Sign
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
SIPOC ReviewDMAIC – Visualize the process
9/16/2015
14
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
SIPOCSuppliers Inputs Process Outputs Customers
Individuals or
organizations
that provide
inputs to the
process.
Material,
information
and/or
services that
are required
by the process
to produce
the outputs
(People,
methods,
machines,
materials &
environment)
The step by
step method
that produces
the output,
defined at a
very high
level- only 5-7
steps
Products,
information,
services
and/or
decisions that
are produced
by the process
Those who
receive the
process
output, pay
for it or are
directly
impacted by
the process
output
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
SIPOC: High Level View
SIPOC• SIPOC: 50,000 Feet View • SIPOC: 50,000 Feet View
Business Process
Map
• Business Process Map: 10,000 Feet View
• Business Process Map: 10,000 Feet View
Standard Work
• Standard Work: 1,000 Feet View• Standard Work: 1,000 Feet View
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
SIPOC – 50,000 Foot View
9/16/2015
15
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Process Mapping: 10,000 Foot View
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Standard Work – 1,000 Foot View
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
SIPOC Example
9/16/2015
16
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: SIPOC
• Review the *Case Study Information*: Scoping Meeting Output
• *Class Discussion*: SIPOC– Volunteer to conduct a SIPOC with IT
Manager and Sponsor for Class
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Process Mapping Review
DMAIC – Visualize the process
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Process Map KeyDifferent functions of the process
Beginning and end points of the process
Any task / activity where work is performed
Places where information is checked against established criteria (standards) & decision made on what to do next
Any time information is waiting before the next process or decision (i.e. in-baskets, out-baskets, waiting to be batched)
Task
Inspect &Decision
Delay
Beginning
& End Points
Function
9/16/2015
17
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Process Map ArrowsUsed between tasks performed by the same person or area, but no physical movement has occurred
Indicates physical movement of information/product from one function to another
Demonstrates electronic movement of information from one person/function to another
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
TIM U WOOD
Transportation
Information/Inventory
MotionWaiting
Over Processing
Over Production
Defect
Underutilization
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Value Added (VA)
• Value Added Activities (VA)-Transforms information into services and products the customer is willing to accept
• VA Activities Must Meet Three Requirements:– Done right the first time– Transformational – Customer is willing to pay for
Typically 1% of a process is Value Added
9/16/2015
18
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Non Value Added (NVA)
Non-Value Added Activities (NVA)• Consumes resources
• Does not directly contribute to service• Customer does not care
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Non Value Added but Necessary (NVAN)
Non-Value Added but Necessary (NVAN)• Customer does not care
• Required to perform the step by current statute or law
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Current State Process Mapping
• In your groups using the Help Desk narrative construct a current state process map
• Identify the TIM U. WOOD• Identify the Value Add Steps
9/16/2015
19
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Discussion*: Project Type
• Based on the information you have received so far – what type of project would be appropriate for the Help Desk Process?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Project Charter Review
DMAIC – Develop Project Charter
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Problem Opportunity Statement
• Describe what the problem is that you are trying to fix (relate back to the customer)
• Define the problem as succinctly as possible:– Who/Where – Who is experiencing the pain, and where
is it being felt? – What – How is the pain being recognized or felt? – When – When did the pain begin? How long has the
problem been around?– Extent – How bad is the problem (systematic issue)?
Tip: Don’t include a solution. If you already know the solution, then Just Do It Project.
9/16/2015
20
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Project Metrics
• Goals of the Projects: Use SMART criteria– Specific– Measurable– Attainable– Resource Requirements/Realistic – Time Boundaries)
• Project Charters are living documents – that should be refined as more information/data is collected
Tip: Examples Include Defect Definition (DPU and %
Defective) or Time Goals (Lead Time and Cycle Time)
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Project Scope
• Clearly define what is “in” scope of the project and what is “out” of scope– Includes Exclude Table
– CT Flowdown
• Identify the beginning and ending steps of the process identified in the “P” of SIPOC
Tip: Make sure the scope is manageable.
If it’s too big, break it down into smaller chunks
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
LeanOhio Boundaries
• No one loses their job because of the Lean event, but duties may be modified
• No additional staff• No additional money• No legislative changes or changes to
collective bargaining agreements• No IT solution until it is determined that an IT
solution is needed• Other?
9/16/2015
21
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Includes/Excludes Tool
Includes Excludes
What
Where
Time
frame
Who
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Includes/Excludes Example
Includes Excludes
Full Time Staff Temp Workers-Consultants
Audit Promotions
Administration Interns
Hired in the past year Hired over 1 year ago
General Services Outside the Agency
All Regions Transfers
HR Process
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
CT Flowdown:DOP Help
Desk Example
DOP IT
9/16/2015
22
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Project Benefits
• Tangible Benefits examples:– Overtime– Postage– Paper– Mailing
• Intangible Benefits – improved morale– time to work on core mission– improved customer satisfaction
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
TRAIL Chart
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Project Charter
• Complete initial project charter/TRAIL chart for the DOP Help Desk Process using the information you have collected so far. – For trail chart you may need to use titles
instead of specific names for SMEs
• *Questions*: Next Steps Questions
9/16/2015
23
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
4 VoicesDMAIC – Identify Critical to Quality
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
4 Voices Review
• Voice of the Customer (VOC)• Voice of the Employee (VOC)
• Voice of the Process (VOP)• Voice of the Business (VOB)
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: 4 Voices
• Identify questions for Employee Focus Group
• Identify questions for Customer Survey• Identify Process Measures (Outputs)
Important to Help Desk• Identify needs of Organization/
Agency• Report out
9/16/2015
24
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: VOC/VOE (Part 2)
• Review *Case Study Information*: VOE Focus Group and Customer Survey
• What conclusions can you draw about the current state process based on the VOE/VOC feedback?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Critical To QualityDMAIC – Identify Critical to Quality
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Process Output Indicators: CTQ’s
VOC - Voice of the CustomerCCR - Critical Customer RequirementsCTQ - Critical to Quality
Delivery Time
Reliability
Cleanliness
Service Level
VOC CCR CTQ
9/16/2015
25
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Translating VOC’s to CCR’s
• Often the Voices of Customers (VOC) are not specific and not in technical language.
• Sometimes the Voices of Customers are stated solutions
• Therefore VOC needs to be translated to the Critical Customer Requirements (CCR)
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Translating VOC to CCRs• A Critical Customer Requirement (CCR) is a
requirement that is important to the customer.
• CCR can be measured (VOC may not be measurable)
• Establishes a target– Customer specifications– Acceptable range of performance
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Translating VOC to CCR: Examples
VOC CCR
I’m always on hold or get
transferred to the wrong person
Customer reaches correct
person the first time within 30
seconds
I need faster serviceI need my deliveries in 3 days or
less
My form keeps getting sent
back
All needed information will be
collected the on the first entry
9/16/2015
26
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Translating CCR’s to CTQ’s
• Often the Critical Customer Requirements are not stated in a way that can be measured internally
• Therefore CCR needs to be translated to the Critical to Quality (CTQ), which can be measured in the agency
For some projects, CCR may be the same as CTQ
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Comparison of VOC,CCR and CTQ
VOC CCR CTQ
LanguageCustomer
language
Technical/
specific
language
Technical/
specific
language
MetricMay not
measurable
Can
measure in
final product
Can measure
in business or
factory
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Translating VOC to CCR to CTQ: Examples
VOC CCR CTQ
I’m always on hold
or get transferred to
the wrong person
Customer reaches
correct person the first
time within 30 seconds
Customer reaches
correct person the
first time within 30
seconds
I need faster serviceI need my deliveries in 3
days or less
Our process lead
time for this product
needs to be 2 days or
less
It takes to long to
get my application
processed
Need decision to
customer in 5 days
Need application
reviewed and
decision in 3 days
9/16/2015
27
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: VOC-CTQ
• Complete VOC, CCR and CTQ for the Help Desk Process based on the Voices feedback
• Update Project Charter
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Discussion*: Next Steps
• Based on the information you have received so far – should the project continue?
• If the project should continue - what type of project would be appropriate for the Help Desk Process?
• What are the team’s next steps?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
MEASURE
9/16/2015
28
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Measure
• Measure Purpose: To determine what to measure, manage the measurement data collection, develop and validate measurement systems and determine process performance.
1. Understand Metrics2. Validate Measurement System3. Determine Process Performance
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Measure Tools• Metrics/Goals• Pareto Chart• Graphical Displays• Operational
Definitions• Process Map• Types of Data• Run Chart• Histogram• Data Integrity Audit
• Attribute Agreement Analysis
• Data Collection Plan• FMEA• C&E Matrix• Decision Tree• Control Charts• Gage R&R• Capability Study
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The Fundamental Principle of Six Sigma
• Uncontrolled process variation leads to poor performance, higher costs, and unpredictable results.– “Variation is evil”
• Variation in process results is produced by variation in the causal factors.– “Y is a Function of X”
• These causes of variation can be identified, measured, prioritized, and then controlled.
9/16/2015
29
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Control Results by Controlling the Causes
• Processes fail to perform whenever the results are not on target– We can be chronically off target– We can be occasionally off target
• Either way, whenever the results are off target it’s because the causes are off target
• Variation from target wastes money, time, and irritates customers
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
How do Statistics Relate to Six Sigma?
• Essentially, the field of statistics deals with the study of variation
• Statistical tools will allow us to quantify the problem, investigate the root causes, develop a solution, and confirm the improvements over time
• Let the Data Take You There!
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
First Lets Take a Moment and Think about Data
• Define your Primary Metric for your Project
• Define your Secondary Metrics
Be intentional when you collect data: Think it through
9/16/2015
30
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Secondary Metrics
Keep Me Honest!
Helps avoid sub-optimization!
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Balancing Measures
• Speed vs. Cost• Time vs. Quality
• Time vs. Cost• Quantity vs. Quality
• Other
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Discussion*
• What are the Primary and Secondary Measures for the Help Desk Process?
9/16/2015
31
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Types of DataDMAIC – Understand Metrics
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Two Basic Types of Data
Continuous (Measured on a continuum)
•Time•Money•Weight•Length•Speed•Temperature
Discrete (Categories)
Ordinal•Count defects•# approved•# of errors•Rank
Nominal•Yes/No•Group A, B..•Good/Bad•Pass/Fail
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
• It’s important to know what type(s) of data you’re collecting… Different types of data require different types of tests
Types of Data
Discrete ContinuousDiscrete Data can be Qualitative
AND QuantitativeContinuous Data is Quantitative
ONLY
Finite Numbers No Fractions Here!
Usually Associated with Measurements & Fractions
Work Here!
Weakest Strongest
9/16/2015
32
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Discrete Data
Can only have the values
2, 3,4,5,6,7,8,9,10,11,and 12
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Continuous Data
HeightTime
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Types of Data• It’s important to know what type(s) of data you’re collecting…
Different types of data require different types of tests
Discrete ContinuousDiscrete Data can be Qualitative
AND QuantitativeContinuous Data is Quantitative
ONLY
Nominal Ordinal Interval Ratio
Finite Numbers No Fractions Here!
Usually Associated with Measurements & Fractions
Work Here!
Weakest Strongest
9/16/2015
33
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Characteristics of Discrete Data
- The weakest level of data. - Made up of values that are distinguished by name only.- There is no standard ordering scheme.
-Stronger than nominal data, but not as strong as Continuous data.- Like nominal data, also made up of values that are rank ordered values or ranked groups.- Different than nominal data because there is an ordering scheme.
Nominal Ordinal
Discrete Data can be Qualitative AND QuantitativeFinite Numbers – No Fractions Here!
Discrete Data
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Discrete: Nominal Data
Nominal: Name Only
You can’t perform arithmetic operations, like addition or subtraction.
Counties
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Discrete: Ordinal Data
Ordinal: Implied Order
May be unable to state whether the intervals between each value
are equal
Likert Scale
9/16/2015
34
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Characteristics of Continuous Data
Interval Ratio
Interval data has an ordering scheme, but unlike ordinal data, the differences between data is meaningful and can be measured. Arithmetic operations are possible. However, interval data lacks a zero starting point, a characteristic unique to Ratio data.
Ratio data has an ordering scheme, has differences that are meaningful, can be measured, and arithmetic operations are possible. Ratio data has a natural starting point and can identify absolute zero.
Continuous DataContinuous Data is Quantitative ONLY
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Continuous: Ratio Data
Ratio: Interval data with a zero starting point.
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Data Summary
• Nominal data used to “name,” or label a series of values
• Ordinal data provide good information about the order of choices, such as in a customer satisfaction survey
• Interval data give us the order of values + the ability to quantify the difference between each one
• Ratio data give us the ultimate–order, interval values, plus the ability to calculate ratios since a “true zero” can be defined
9/16/2015
35
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Data Summary �
Provides: Nominal Ordinal Interval Ratio
“Counts,” aka “Frequency of
Distribution”� � � �
Mode, Median � � �
The “order” of values is
known� � �
Can quantify the difference
between each value� �
Can add or subtract values � �
Can multiple and divide
values�
Has “true zero” �
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Questions*: Types of Data
• Identify the type of data from the Help Desk Data Set
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Data Collection PlanDMAIC – Understand Metrics
9/16/2015
36
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Develop a Measurement Plan -Types of Data
• Before data collection starts, classify the data into different types: continuous or discrete
• This is important because it will:– Provide a choice of data display and analysis
tools– Dictate sample size calculation– Provide performance or cause information– Determine the appropriate control chart to use– Determine the appropriate method for calculation
of Sigma Level
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Data Collection Method
Measurement management starts with a data collection methodology
Identify
Measures
Step 1
Develop operational definitions
for measure
Step 2
Develop measurement plan
Step 3
Collect data
Step 4
Display and evaluate data
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Step 1: Operational Definition
•A precise description of the:– Specific criteria used for the measures (the what)– The methodology to collect the data (the how)– The amount of data to collect (how much)– Responsibility to collect the data (the who)
Why do you need to define “Operational Definitions”
9/16/2015
37
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Six Sigma and Operational Definitions
Example: Number of Days to complete an Activity• Business Days• Calendar Days• Holidays • Employee Leave
• Each of these cases may require a very different approach for gathering the data.
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Step 2: Develop a Measurement Plan
Determining current process performance usually requires the collection of data. When developing a measurement plan ensure that:
– The data collected is meaningful
– The data collected is valid
– All relevant data is collected concurrently
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Questions to Answer
• What precise data will be collected?
• Do we analyze all relevant data or a sample?
• What tools are necessary?
• What logistical issues are relevant?
• What do you want to do with the data?
9/16/2015
38
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Samples Vs. Populations
DMAIC – Understand Metrics
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Sample vs. Population
• A population is the collection of people, items, or events about which we want to draw conclusions.
• A sample is a subset of the population.
• If selected properly, the sample will display characteristics similar to the parent population
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Why Sample
• It is impractical or sometimes impossible to take measurements on the whole population– Populations tend to be large, therefore it can be costly
to use all data– For processes, the total population may not yet exist
• If it is practical and feasible to get data for the entire population - then do it!
9/16/2015
39
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Statistical Inference
• A well-selected sample represents all of the key characteristics of the population from which it is drawn
• Statistical inference enables us to draw a conclusions about an entire population using only sample data
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Sampling Error
• When choosing your sample beware of “sampling error”
• We will use sampling methods to help avoid sampling error– Random– Systematic– Stratified
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Sample Calculator - Continuous Data
9/16/2015
40
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Central Limit Theorem
• When studying averages, as the sample size gets larger, the distribution of averages more closely approximates normality
• The average of the sample averages is approximately equal to the population average
• The standard deviation of sample averages is approximately equal to the population standard deviation divided by the square root of sample size (n)
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Confidence Intervals• Confidence intervals give an estimated range of values
which is likely to include an unknown population parameter, the estimated range being calculated from a given set of sample data.
• Statistics such as the mean and standard deviation are only estimates of the population mean and standard deviation, and are based on one sample only.
• Because there is variability in these estimates from sample to sample, we can quantify our uncertainty using statistically based Confidence Intervals (C.I.’s)
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov120
Sample Data Measurement Plan Form
Performance Measure
Operational Definition
Data Source
and Location Sample Size
Who Will Collect the
Data
When Will the Data Be
Collected
How Will the Data Be
Collected
Other Data that Should Be
Collected at the Same Time
How will the data be used? How will the data be displayed?
Examples:� Identification of Largest Contributors� Identifying if Data is Normally Distributed� Identifying Sigma Level and Variation� Root Cause Analysis� Correlation Analysis
Examples:� Pareto Chart� Histogram� Control Chart� Scatter Diagrams
9/16/2015
41
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity* Data Collection Plan
• Create a Data Collection Plan for the DOP Help Desk Process
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Graphical DisplaysDMAIC – Understand Metrics
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
9/16/2015
42
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Basic Graphical Displays
• Bar Chart compares data across categories
• Pie Chart compares data across categories and the proportion of each category to the whole
• Histogram displays the distribution of the data by summarizing the frequency of data values within each interval
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Baseline DataDMAIC – Understand Metrics
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Baseline data
• Baseline Data is the data before you have made any changes
• This is the Current State of the Process• Used to compare to data obtained after
implementation (or changes are made to your process) to show– Are we doing better– Are we doing the same– Are we doing worse
9/16/2015
43
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Baseline Data
• Review baseline data and create visuals of data
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Discussion*: Baseline Data
• What initial information can you conclude based on this spreadsheet?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Types of Descriptive Statistics
DMAIC – Understand Metrics
9/16/2015
44
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Types of Basic Descriptive Statistics
The Key Characteristics of
Data
• Center
• Spread
• Shape
• Stability over Time
Statistics for Discrete Data
• Defectives (Units)
• Statistics for Defects (Errors)
Statistics for Continuous Data
• T-Test
• ANOVA
• Regression
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Data varies across its measurement scale
• Except in the rarest of circumstances, data will vary…even when nothing in the process seems to be changing
– Truth: Something is changing to cause the variation, we just may not be able to see what/how is changing
• Any set of data will have values that distribute across the measurement scale
– This is cleverly called a data distribution, or simply “distribution”
• Knowing the data type and distribution is critical to choosing the right statistical tools
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Foundation Statistics
• Statistics for Variable/Continuous Data
– Describing the process center
– Describing the process spread
– Describing the process shape
– Describing the process stability
• Statistics for Defectives (Units)
• Statistics for Defects (Errors)
9/16/2015
45
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Descriptive Statistics
We look at a variety of Descriptive Statistics to provide insights into different aspects of data –they provide summary information that we can’t see in graph or a table of data values (for example, range, mean and median)
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Four Characteristics of Data
DMAIC – Understand Metrics
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The Key Characteristics of a Distribution
• Where on the measure scale does the data appear to gather or “clump”?– What is the center of the data?
• How does the data distribute around the center?– What is the spread of the data?
• What values are more frequent and less frequent?– What is the shape of the data?
• How do the above characteristics behave over time?– What is the stability of the data?
9/16/2015
46
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The Key Characteristics of a Distribution
• What is the center of the data?
Where on the measure scale does the data appear to gather or “clump”?
• What is the spread of the data?
How does the data distribute around
the center?
• What is the shape of the data?
What values are more frequent and
less frequent?
• What is the stability of the data?
How do the above characteristics
behave over time?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Four Key Characteristics of Continuous Data
Center
• Mean
• Median
• Mode
Spread (Variation)
• Range
• Standard Deviation
• Variance
Shape
• Skew
• Kurtosis
Stability
• Run Charts
• Control Charts
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Measures of Center
• The mathematical average of a set of data point values. (Sum of all data points/number of data points)
Mean
• The middle data point when the data is sorted by value, where 50% of the observed values are below and 50% are above. If there is an even number of data points, then average the two points in the middle
Median
• The most frequently occurring data point valuesMode
9/16/2015
47
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Four Key Characteristics of Continuous Data
Center
• Mean
• Median
• Mode
Spread (Variation)
• Range
• Standard Deviation
• Variance
Shape
• Skew
• Kurtosis
Stability
• Run Charts
• Control Charts
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Spread: Variation
• Accuracy vs. Precision
• Measures of Variation
– Range
– Standard Deviation
– Variance
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Accuracy vs. Precision
Accuracy describes
Centering
How close to
target?
Precision describes
Spread.
How close
together?
9/16/2015
48
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Range
• Range: This is the difference between the largest and the smallest data point values.
• The purpose is to measure the dispersion (range) between the highest and lowest values of a data set.
Range = Maximum Value - Minimum Value
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Standard Deviation
• Deviation: This is the distance between a data point value and the mean.
• These deviations for each data point will be used to calculate and describe the variation in a set of data.
( )XXDeviation −=
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov144
Standard Deviation
• Deviation = distance from mean
9/16/2015
49
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Standard Deviation
• Standard Deviation: It is a measure of the average dispersion about the mean.
• Population Standard Deviation (σ):
( )N
N
iix∑
=
−= 1
2µσ
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Variance
• Variance: The variance is the square of the standard deviation, σ2.
• For the sum or difference of two independent random variables, the variance is the sum of the individual variances.
σa2 = Variance of Variable a
σb2 = Variance of Variable b
σtotal2 = σ2
a + σ2b = Total Variance of a and b
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Four Key Characteristics of Continuous Data
Center
• Mean
• Median
• Mode
Spread (Variation)
• Range
• Standard Deviation
• Variance
Shape
• Skew
• Kurtosis
Stability
• Run Charts
• Control Charts
9/16/2015
50
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Evaluating Shape
• Graphical Methods
– Dot Plots
– Histograms
• Quantitative Methods
– Skewness and Kurtosis
– Goodness of Fit
– Normality Tests
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Probability Distributions
• Probability Distribution: This is the tendency of a large numbers of observations from a process to group themselves around some central value with a certain amount of variation or “scatter” on either side
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Probability Distributions
• There are many types of probability distributions.– Binomial, Poisson, Uniform, Normal, Beta,
Exponential, Weibull, Gamma, etc.
• Probability distributions are either discrete or continuous. It depends on the random variable
9/16/2015
51
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Normal Distribution
• The Normal Distribution (Gaussian Distribution) is the class bell-shaped curve which approximately describes many phenomenon in industry and science, and it is always the distribution of sample means from any distribution
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The Normal Distribution
• The “Normal” Distribution is a distribution of data which has certain consistent properties
• These properties are very useful in our understanding of the characteristics of the underlying process from which the data were obtained
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The Normal Curve and Probability
43210-1-2-3-4
40%
30%
20%
10%
0%
95%
Pro
ba
bil
ity
of
sam
ple
va
lue
Number of standard deviations from the mean
► The area under the curve can be used to estimate the probability of a certain measurement value occurring
99.73%
68% Cumulative
probability of
obtaining a value
between two values
9/16/2015
52
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Skewness - Kurtosis
• Skewness refers to a lack of symmetry. A distribution is skewed if one tail extends farther than the other
• Kurtosis refers to how sharply peaked a distribution is
• A value for kurtosis and skew is included with the graphical summary
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Four Key Characteristics of Continuous Data
Center
• Mean
• Median
• Mode
Spread (Variation)
• Range
• Standard Deviation
• Variance
Shape
• Skew
• Kurtosis
Stability
• Run Charts
• Control Charts
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Stability
• Is the process changing over time?• Are there any trends, clusters, oscillations,
etc?• Run Charts and Control Charts will help
determine if the process is stable• A stable process is a process which is free
of assignable causes (in statistical control)
9/16/2015
53
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov157
Foundation Statistics
• Statistics for Variable/Continuous Data– Describing the process center– Describing the process spread– Describing the process shape– Describing the process stability
• Statistics for Defectives (Units)• Statistics for Defects (Errors)
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Defective vs. Defect
• Defective: Good/Bad, Yes/No• Defect Data: Used where more than one
defect per unit
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Statistics for Defective (Yield) Data
Defective data: Good/bad, Go/No-go, on/off
Proportion Defective p = (# Defective)/nin a sample of n
Average Proportion
Defective
pTotal # Defective
=Total n
BAD
GOOD
9/16/2015
54
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Foundation Statistics
• Statistics for Variable/Continuous Data
– Describing the process center
– Describing the process spread
– Describing the process shape
– Describing the process stability
• Statistics for Defectives (Units)
• Statistics for Defects (Errors)
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
DPU, DPO and DPMO
• Metrics that express how your product or process is performing, based on the number of defects
• Choosing the appropriate quality metric helps you assess performance against customer expectation
• Can also develop project baselines and improvement goals
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Defects per Units (DPU)
• Defects per unit (DPU) is the number of defects in a sample divided by the number of units sampled.
• Used where more than one defect per unit is likely
Average Defects per Unit (DPU)
cTotal # of Defects
=Total # of Units
Inspected
9/16/2015
55
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
DPU Example• Your printing business prints custom orders. Each order
is considered a unit. Fifty orders are randomly selected and inspected and the following defects are found:– Two orders are incomplete – One order is both damaged and incorrect (2 defects) – Three orders have typos
• Six of the orders have problems and there are a total of 7 defects out of the 50 orders sampled
• DPU = 7/50 = 0.14
• On average, this is your quality level and each unit of product on average contains this number of defects
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Defects per opportunity (DPO)
• Defects per opportunity (DPO) is the number of defects in a sample divided by the total number of defect opportunities
Average Defects per Opportunity
(DPO)
Total # of Defects=
Total # of Defect Opportunities
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
DPO Example• Each custom stationary order could have four
defects (incorrect, typo, damaged, or incomplete)
• Each order has four defect opportunities
• Fifty orders are randomly selected and inspected and the following defects are found:– Two orders are incomplete
– One order is both damaged and incorrect (2 defects)
– Three orders have typos
• Six of the orders have problems, and there are a total of 7 defects out of the 200 opportunities (50 units * 4 opportunities / unit)
• DPO = 7/200 = 0.035
9/16/2015
56
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Defects per million opportunities (DPMO)
• Number of defects in a sample divided by the total number of defect opportunities multiplied by 1 million.
• DPMO standardizes the number of defects at the opportunity level and is useful because you can compare processes with different complexities.
Average Defects per Million
Opportunities (DPMO)
Total # of Defects=
Total # of Defect Opportunities
X 1,000,000
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
DPMO Example• Each custom stationary order could have four defects -
incorrect, typo, damaged, or incomplete.
• Fifty orders are randomly selected and inspected and the following defects are found:– Two orders are incomplete
– One order is both damaged and incorrect (2 defects)
– Three orders have typos
• There are a total of 7 defects out of the 200 opportunities
• Therefore, DPO = 0.035 and DPMO = 0.035 * 1000000 = 35,000
• If your process remains at this defect rate over the time it takes to produce 1,000,000 orders, it will generate 35,000 defects
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Descriptive Statistics
• Open Help Desk Baseline Data in Excel and conduct Data Analysis for Time to Process Data
9/16/2015
57
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Discussion*: 4 Characteristics
• What is the Range of our Baseline Help Desk Process?
• How is the help desk doing?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Normal DataDMAIC – Understand Metrics
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Determining Normality
• Not all data is normal– Belts must know whether or not the data is
normal as different tests apply in different circumstances
• Normal data is defined as data that has “normal” variation
9/16/2015
58
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Probability Distributions
Normal Data – Why should I care!• Normally distributed data exhibit
predictable traits and probabilities• In practice, we are frequently confronted
with data that is not normal• The first step to take is to look at how the
data is distributed
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Distributions
When we measure a quantity in a large number of individuals we call the pattern of values obtained a distribution
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
What is a distribution?A way of describing and viewing the data that you have
– Focuses on:• Shape of the data• ‘Range’ of the data e.g. maximum and minimum
point
9/16/2015
59
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Normal Data
• In this context the name “normal” causes much confusion. In statistics it is just a name
• Indeed, in some arenas normal distributions are rare
• Various methods of analysis make assumptions about normality, including:– correlation, regression, t tests, and analysis of
variance
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Parametric Testing• Tests like the t-test assume normality and are
sometimes called “parametric tests”.– “Parametric” implies that a distribution (shape)
is assumed for the population - commonly, and in this case, the Normal Distribution
• Advantage of a parametric test is higher statistical power if the assumptions hold
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Nonparametric Testing• Nonparametric tests can give you more flexibility
because they do not assume a shape. This allows you to use them with any data.
• However, even Nonparametric tests usually have some minimal shape requirements– Must be unimodal– Each group being tested must have same
general shape (for a two variable test)
9/16/2015
60
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
4 steps to Normality!• Plot your data: Histogram (eye ball it)• Compare Mean and Median (if similar
it suggests normal data)• Look at Skew and Kurtosis (if both
close to zero it suggests normal data)• Tests for normality (Anderson-
Darling test)
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
4 Steps: Hypothesis Tests
• Anderson-Darling test is a hypothesis test • It is the hypothesis test we run before we
run our other hypothesis tests• It tells us which test to run: normal or non-
normal
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Intro to Hypothesis Testing
DMAIC
9/16/2015
61
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Hypothesis testing assumes a condition exists in a population and a sample
is taken to confirm or deny the assumption.
Hypothesis Testing
• Helps to properly handle uncertainty• Minimizes subjectivity
• Prevents the omission of important information
• Manages the risk of decision errors
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
In hypothesis testing we first establish the null hypothesis (Ho) , the
assumption, and an alternative hypothesis (Ha) .
Hypothesis Testing• Ho = Null hypothesis - always specifies a value for
the population parameter. We assume the null hypothesis is true
• Ha = Alternative hypothesis - answers the question by specifying that the parameter is one of the following– Greater than the value shown in the null hypothesis– Less than the value shown in the null hypothesis– Different from the value shown in the null hypothesis
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
State a “Null Hypothesis” (Ho)
Gather evidence (a sample of reality)
DECIDE:
What does the evidence suggest?
Reject Ho? or Don’t Reject Ho?
Hypothesis Testing Steps
9/16/2015
62
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Hypothesis Testing of Means
• Ho: Mean of Group A = Mean of Group B • Ha: Mean of Group A ≠ Mean of Group B • Example:
– Ho: Mean time to hire an employee in Region 4 is the same mean time to hire an employee in Region 6.
– Ha: The mean time it takes to hire an employee is different in Region 4 and Region 6.
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Hypothesis Testing of Variances
• Ho: Variance Group A = Variance Group B• Ha: Variance Group A ≠ Variance Group B• Example:
– Ho: Time to serve variance at the Region 4 and Region 6 is the same
– Ha: Time to serve variance is different at the Region 4 and Region 6
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Examples of Hypothesis Testing
• It is assumed the hiring processes for Region 4 and Region 6 are equal
• The null hypothesis (Ho) is that they are “equal”• We must collect data to show that it is statistically
different• If the data indicates there is a statistical difference,
we find that the hiring processes are not equal
9/16/2015
63
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Hypothesis and Decision Risk
• There is a known degree of risk and confidence when accepting or rejecting a hypothesis
– Alpha risk ( α): The rejection of the null hypothesis when it is true.
• Type I error- P Value
– Beta risk ( β): The acceptance of the null hypothesis when it is false.
• Type II error- missed a factor
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Your Decision
Accept Ho
The
Truth
Ho True
Ho False
Type I
Error
(α-Risk)
Type II Error
(β -Risk)
Correct
Correct
Reject Ho
Decision Errors
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
P-Values
• The p-value is the probability that such error could occur when Ho is true
• The p-value is based on an assumed or actual reference distribution in tests such as:– Normality tests
– Chi-Square
– Descriptive Statistics
– t-distribution
– F-distribution
• Every test of significance is a test of a null hypothesis
9/16/2015
64
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
What are P-values Used For?
• Small P-Value
• Ample Evidence
• Ho is Rejected
• Support alternate
• Large P-Value
• Little evidence
• Ho is Not Rejected
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
How Low Must P be?
• P-value is the probability that the null hypothesis is true.
MANTRA
“If p is low the Ho must go”
• 1- p measures our confidence in the alternative hypothesis
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
More on P values
• The p-value most industries use is – p-value = 5% (α = .05) – there is less than a 5% chance that these
observations could have occurred randomly
9/16/2015
65
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Summary
• First step of hypothesis testing is to decide on your hypothesis
• We assume the null hypothesis is true• We take a sample and test for our hypothesis• We determine the risk in accepting or rejecting the
null hypothesis• P-value of a test of hypothesis is the smallest (.05
normally) value of alpha that would lead to rejection of the null hypothesis
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Discussion*: Normality Hypothesis Test
Write your hypothesis for Normality: Help Desk
Time to Process Tickets
• Ho
• Ha
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Minitab Demonstration*: Normality Hypothesis Test
• Exercise: Calculate normality for Help Desk data: Time to Process
• Mantra: Stat-Basic Stat-Graphical Summary
9/16/2015
66
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Discussion*: Interpret Normality Test
Interpret the Normality Test
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Run ChartsDMAIC
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Run Charts
• Data graphed in chronological order• Run charts detect for trends, clustering,
mixtures, or oscillations in the process• Run charts can be used with non-normal
data!
9/16/2015
67
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Run Chart
• In Excel create and interpret a Run Chart using the data provided below for the Help Desk Ticket - July 2014 data
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Control ChartsDMAIC
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Basic Control Charts
What is the tool?• A Control Chart can be considered a road map of…
– ... where you have been.– ... where you are– ... where you may be headed
• Because of the statistics, control charts can recognize good and bad changes
9/16/2015
68
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Why use Control Charts
• Detect changes in a process: when something is statistically different in my process
• Obtain a basic understanding of when a process is “out of statistical control”
• To visually see how your process varies within the control limits
• Know when to “freak out”
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
* A process is Stable, Predictable, and In-Control when only
Common Cause Variation exists in the process
Types of Variation –Common vs. Special
• Common Cause (Noise)*– Is present in every process – Is produced by the process itself (the way we
do business)– Can be removed and/or lessened but requires a
fundamental change in the process
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
* A process exhibiting Special Cause variation is said to be
Out-of-Control and Unstable
• Special Cause (Signals)*– Unpredictable– Typically large in comparison to Common
Cause variation– Caused by unique disturbances or a series of
them– Can be removed/lessened by basic process
control and monitoring
Types of Variation –Common vs. Special
9/16/2015
69
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
1 Sigma
2 Sigma
3 Sigma
1 Sigma
2 Sigma
3 Sigma
68%
95%
99.7%
% of Data PointsUCL
LCL
TIME
The ItemWe Are
Measuring
Rules of Standard Deviation“Where does the data lie?”
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Data Plotted Over Time
MO
NIT
OR
ED
C
HA
RA
CT
ER
IST
IC
UCL
Center Line
LCL
UCL = Upper Control Limit / LCL = Lowe r Control Limit
Plotted Data
“Over Time” means in chronological order
The Basic Control Chart - Key Elements
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The Rules We Will Most Often Use:
Rule 1: One point more than 3 sigmas from center line
Rule 2: Nine points in a row on the same side of center line
Rule 3: Two out of three points more than 2 sigmas from center line
(same side)
Rule 4: Four out of five points more than 1 sigma from center line (same
side)
Pattern Rule: A pattern repeats itself
Control Chart Rules
When one of these rules is broken, we say that the process is “out
of statistical control”
9/16/2015
70
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Rules of Detection
3σ2σ1σ
1σ
3σ2σ
x
UCL
LCL
1
3σ2σ1σ
1σ
3σ2σ
x
UCL
LCL
3
3σ2σ1σ
1σ
3σ2σ
x
UCL
LCL
4
3σ2σ1σ
1σ
3σ2σ
x
UCL
LCL
2
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Basic Control Charts - Limitations
• Control charts will not pinpoint what or why something has changed
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Control Limits vs. Spec Limits
• Where do Control Limits come from?
• Where do Specification Limits come from?
21
0
9/16/2015
71
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Upper Control Limits = UCLLower Control Limits = LCL
Upper Specification Limits = USLLower Specification Limits = LSL
Is The Process Below Making Defects ?
UCL
LCL
TIME
USL
LSL
Control Limits vs. Spec Limits
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Upper Control Limits = UCLLower Control Limits = LCL
Upper Specification Limits = USLLower Specification Limits = LSL
Is The Process Below Making Defects ?
UCL
LCL
USL
LSL
TIME
Control Limits vs. Spec Limits
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Three Big Control Chart Errors
• Putting specification limits on a Control Chart
• Treating UCL and LCL as specification limits
• Not putting data in chronological order
9/16/2015
72
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Control Charts
When you might use Control Charts
– Measure phase: to separate common cause variation from special cause variation
– Analyze & Improve phase(s): to ensure process stability before completing a hypothesis test (more on that later)
– Control phase: to verify the process remains in control after removing special cause variation sources you identified
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Types of Control Charts Found in Minitab
• Variables Charts (Continuous Data)– I and MR Chart– X-Bar / R Chart
• Attribute Charts (Discrete Data)– np-Chart– p-Chart– c-Chart– u-Chart
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Basic Control Chart Rules• Control charts require maintenance and should be used
sparingly
• The chart type selected must fit this characteristic
• At least 10 data points (samples of parts) must be gathered prior to building these control charts
• Appropriate action must be taken when signaled by the control chart
9/16/2015
73
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
We will look at I-MR, X-Bar R, u and p charts
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Individuals, Moving Range Chart – IMR
• Continuous data• Subgroup size of 1
Used when:• Each measurement needed• Each measurement represents a batch
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*MINITAB Demonstration*: Control Chart
• Create a Control Chart for the Help Desk Time to Process
9/16/2015
74
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Discussion*: Control Chart
• Is our process in Control? - What can we do if our process is out of Control?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Discrete/Attribute Control Charts Defects vs. Defectives
• A defective is an item in a sample that has one or more nonconformance(s) to the specified criteria. The data can take only one of two values, such as pass/fail or go/no go.
• A defect is each nonconformance to the specified acceptance criteria. More than one defect may exist on the unit.
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Discrete Control Charts
Underlying
Sample
Constant
Underlying
Sample
Changes
Defective
(Is/Is Not)
Defects
(How Many Per)
np c
p u
9/16/2015
75
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Calculation of New Control Limits
When are new control limits calculated?
– New control limits are usually recalculated when a there is evidence that the process has been changed and stabilized at a new level.
– They can be recalculated earlier but it is best to have at least 20 subgroups of data available. Remember, control limits Indicate the capability of the process
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Ishikawa DMAIC – Identify Potential Sources
of Variation
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Fishbone Diagram: What
• A picture of various system elements that may contribute to a problem
• Allows the organization of large amounts of information about the problem and its possible causes
• Creates a snapshot of collective knowledge about the problem
9/16/2015
76
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Fishbone Diagram: Why• To identify possible causes of a problem• To identify the most likely cause and
discover root causes• To ensure that all perspectives are looked
at and nothing is overlooked
• To preclude jumping to solutions• To move from opinion to testable theories
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Fishbone Diagram: How (7M’s)
Problem
Statement
MachineMan
Management
Mother Nature
Measurements
Materials Methods
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Fishbone Diagram: How (4P’s)
Problem Statement
PeoplePolicies
Procedures Place
9/16/2015
77
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Fishbone Diagram: Design
Problem Statement
PeoplePolicies
Procedures Place
Primary Cause
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Fishbone Diagram
• Based on the information you have collected so far complete a Fishbone Diagram for the Help Desk Process
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
FMEADMAIC – Identify Sources of
Variation
9/16/2015
78
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Control and Cost
1. Prevention $
2. Detection $$
3. Correction $$$$
– Exponentially expensive
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Failure Mode and Effect Analysis (FMEA)
• Start by listing ways the project might potentially fail (Problem)
• Evaluate the severity (S) of each problem– “1” represents failure with no effect and “10”
represents very severe and catastrophic failure
• Estimate the likelihood (L) of each problem occurring– “1” indicates that failure is rather remote and not likely
to occur and “10” indicates that failure is almost certain to occur
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
FMEA
Severity x Likelihood x Dectectability = RPN
• Estimate ability to detect each problem (D)– “1” is used when monitoring and control systems are almost certain
to detect the failure and “10” where it is virtually certain the failure will not be detected
• Next calculate the Risk Priority Number (RPN)– Multiply S, L and D together
• Sort potential failures by their RPNs and focus on the highest RPNs
• Finally, consider ways of reducing the risk associated with failures with high RPNs
9/16/2015
79
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Rating Severity of Effect Likelihood of Occurrence Abi lity to Detect
10 CatastrophicVery high:
Can not detect
9 Very SevereFailure is almost inevitable
Very remote chance of detection
8 SevereHigh:
Remote chance of detection
7 Moderately severeRepeated failures
Very low chance of detection
6 Loss of funtionalityModerate:
Low chance of detection
5 Significant effectOccasional failures
Moderate chance of detection
4 Obvious effectModerately high chance of
detection
3 Noticeable effectLow:
High chance of detection
2 Minor effectRelatively few failures
Very high chance of detection
1 No effect Remote: Failure is unlikely Almost certain detection
Sample Rating Scale
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Now What?
• Focus on High RPNs• Look for causes• First-find prevention• Secondly-examine detectability• Third-consider correction• Any new processes should consider FMEA
Monitors, Measures, Poka Yoke, Control
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: FMEA
• Based on the information you have collected so far complete a FMEA for Help Desk Process
9/16/2015
80
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Update Project Charter
• Review the Project Charter and make updates if appropriate
• Identify the team’s next steps
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
ANALYZE
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Analyze
• Analyze Purpose: To determine the root causes, estimate population parameters with confidence intervals and to construct hypothesis about the data and test them to determine significance.
1. Identify Sources of Variation 2. Identify/Verify Root Causes3. Characterize and Determine Significant
X’s
9/16/2015
81
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Analyze Tools
• Cause-Effect Diagram
• Pareto Chart• Graphical Displays• Team Norms
• Hypothesis Testing
• Scatter Plot• Design of
Experiments• Value Stream
Mapping
• Multi-Variable Testing
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Pareto ChartDMAIC – Identify Potential Sources
of Variation
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Pareto Chart
• Pareto Chart is a fancy bar graph.
A Pareto chart is a bar graph with the bars sorted in order of decreasing
frequency. It is used to identify the largest opportunity for improvement.
9/16/2015
82
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Pareto Chart: When to Use
• When analyzing data about the frequency of problems or causes in a process
• When there are many problems or causes and you want to focus on the most significant
• When analyzing broad causes by looking at their specific components
• When communicating with others about your data
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Pareto Chart Example
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Pareto Chart
• Create a Pareto Chart of the Type of Ticket for the Help Desk Baseline Data (Flip chart paper)
9/16/2015
83
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Classifying Variables
DMAIC – Characterize the X’s
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Understanding Inputs & Outputs
• Y’s (outputs) are impacted by X’s (inputs)
• Through use of analytical methods, we will understand how various inputs cause the process (outputs) to behave.
– First we must identify the inputs and outputs that are key to the process
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Output Variables• Products, services, or information can all be outputs of a
process
– The measurable “results” of the process are referred to as output variables
– Why “variables”? Outputs change as a result of process variation; we are measuring the key characteristics of the outputs
– Measuring variability is key since outputs often become the inputs to another process (sub-op)
• Output variables = “Y” variables
9/16/2015
84
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Input Variables
• Factors that affect a process and the process’ outputs are called inputs
– The measurable “inputs” on the cause side of the process are called input variables
– When inputs change, it can cause the process to change (ultimately impacting the outputs)
– Inputs for one process are often the results (outputs) from another process
• Input variables = “X” variables
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Classifying Input Variables
Classifications for input variables include:
Controllable (C):• can be changed to see the effect on Y; “Knob” Variables
Uncontrollable (U):• impact the Y but are difficult or impossible to control; ex:
environmental variables such as humidity;
• may be controllable by something/ someone else (output of another process); “Noise” variables
Critical Key Inputs (K):• statistically shown* to have a major impact on the
variability of the Y;
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Classifying Variables
• Identify the Y’s and X’s for the Help Desk Process
• Brainstorm the X’s for the Help Desk Process
9/16/2015
85
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Capability AnalysisDMAIC
Determine Performance Capability
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Why do we do Process Performance Calculations?
• Document baseline performance
• Compare performance before and after
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Process Capability Study
• A process capability is part of an overall strategy of Six Sigma and process improvement that has three objectives:
– Obtain Stable processes– Reduce the Variability of key process outputs– Improve the Capability of key processes through the
reduction of variation and the centering of the process on its target value
9/16/2015
86
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Process Capability Study
A process capability study generally consists of four steps:
Step 1. Verify that the process is stable
Step 2. Determine if the data distribution is normal
Step 3. Calculate the Capability Indices Cp and Cpk; determine Sigma Quality Level
Step 4. Make recommendations for process improvement
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Voice of The Customer
USLLSL
Capability Studies Compare VOC to VOP
Voice of The Process
• We continually gather data on our processes and ask:– “Is it capable of producing defect free products/services?”
• Gathering and analyzing the data in response to this question is called a Capability Study
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Specification Limits
Where do Specification Limits come from?• Customer Needs• Laws/Regulations• Benchmarking• LeadershipWhere do Control Limits come from?• The DATA
9/16/2015
87
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
“Defects”(Χ ≥ 4)
1 2 3 4 5 6 7 Number of Mistakes
“Defects”(x > 130 Min)“Non-Defects”
(x < 130 Min)
15 110 115 120 125 130 135 140
Assembly Time (Minutes)
“Defect Free”(Χ ≤ 3 )
CustomerRequirement
CustomerRequirement
Discrete Continuous
How do you look at your data?
Classifying Data
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Capability Indices
• Capability indices are ratios of the process spread and specification spread
• They are unitless values so that you can use them to compare the capability of different processes
• Many practitioners consider 1.33 to be a minimum acceptable value for capability indices; and most practitioners believe a value less than 1 is not acceptable
• In general, the higher your values, the more capable your process
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Centering – Put The Process On Target
Cpk: Short Term – Ppk: Long Term
Spread – Reduce The Variation
Cp: Short Term - Pp: Long term
LSL USL
DefectsDefects
Process Capability - The Strategy
9/16/2015
88
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Very Low Probability of
Defects
Very Low Probability of
Defects
LSL USL
BetterProcess Capability After
Very High Probability of
Defects
Very High Probability of
Defects
LSL USL
Poor Process Capability Before
Spread Reduction
Very Low Probability of
Defects
Very Low Probability of
Defects
LSL USL
BetterProcess Capability AfterVery High
Probability of Defects
LSL USL
Poor Process Capability Before
Mean Shift Needed
The Focus of Improvement
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Process Capability Ratios
• Statisticians developed 2 key measures for
capability
USLC
- LSL|
6sp =
Cpk = Min (X - LSL
3s
USL - X
3s, )
Total Tolerance
Process Spread=
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Interpretation of Cp
LSL USL
Cp
<1.0
1.0 – 1.5
>1.5
>2.0
Poor capability
Marginal
capability
Good
capability
6 Sigma
level
capability
Interpretation
9/16/2015
89
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Cp vs. Cpk
• Disadvantage of the Cp and Pp indices is that they compare only the process spread with the specification spread, but do not describe how far the process is from the target region
• Assess the amount of variation, but not the accuracy of the process
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Cp vs. Cpk• Use graphs in combination with other
capability indices (such as Cpk, Ppk,) to more thoroughly evaluate the performance of your process
• Cpk, CPL and CPU indices measure the distance between the process average and the specification limits, compared to the process spread
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
CPL, CPU and Cpk• CPL: Measures how close the process
mean is running to the lower specification limit
• CPU: Measures how close the process mean is running to the upper specification limit
• Cpk: Equals the lesser of CPU and CPL• Many industries use a benchmark value of
1.33
9/16/2015
90
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
CPL, CPU and Cpk• If Cpk, CPU, and CPL are equal, the
process is centered at the exact midpoint of the specification limits
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
CPL, CPU and Cpk• If CPL does not approximately equal CPU,
the process is not centered• When CPL < CPU, the process is more
likely to produce defective units that violate the lower specification limit
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
CPL, CPU and Cpk
• When CPU < CPL, the process is more likely to produce defective units that violate the upper specification limit.
9/16/2015
91
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
T USLLSL
Cpk = 2.0 Cpk = 1.0Cpk = 1.0
Three Processes with Cp = 2.0
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Is The Process
In Control ?
Is It Producing
Defects ?
Short vs. Long Term Capability
• Cp, Cpk: A Short-term Capability study covers a relatively short period of time (Days, Weeks) generally consisting of 30 to 50 data points. The actual number depends on the subject under study
• Pp, Ppk: A Long-term capability study covers a relatively long period of time (Weeks, Months) generally consisting of 100-200 data points. Again, the actual amount depends on the subject under study
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
LSL
Short-Term
Capability
Long-Term
Capability
Over time, a process tends to shift by approximately 1.5σ
Short-Term
Capability
USL
The Dynamic Process
9/16/2015
92
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Determining Spec Limits
• Where do Specification Limits Come From?
• Where do Control Limits Come From?
• Based on the VOC-CTQ exercise determine the Help Desk Process Lower Specification and Upper Specification Limits
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Minitab Demonstration*: Capability Analysis
• Minitab Demonstration: Conduct Capability Analysis of the Help Desk Time to Process Tickets using the defined Lower and Upper Specification
• Is the current help desk ticketing process capable?
• What are the team’s next steps?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Catapult Capability Exercise
• Set up catapult with the target being 70”
• Take 30 shots – 2 partners each
• Use the following specifications (customer
expects you to be between 64 and 76)
– LSL = 64”
– USL = 76”
• Determine Capability
9/16/2015
93
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
1 Variable TestingDMAIC – Understand Metrics
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Types of Data• It’s important to know what type(s) of data you’re collecting…
Different types of data require different types of tests
Discrete ContinuousDiscrete Data can be Qualitative
AND QuantitativeContinuous Data is Quantitative
ONLY
Nominal Ordinal Interval Ratio
Finite Numbers No Fractions Here!
Usually Associated with Measurements & Fractions
Work Here!
Weakest Strongest
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Think First, then Plot The Dots!
• Pictures are worth a thousand words • By way of an analogy, even though you
could statistically prove that people are heavier before a haircut than they are afterwards, nobody seeks out a barber shop when they go on a diet
• Viewing the data graphically helps to confirm how much practical significance exists in what you’ve found
9/16/2015
94
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov280
“1-Variable” Testing
• “1-Variable” testing involves evaluating 1 column of data at a time.
• Your project will require you to collect data on a number of different variables
• With this class of tools, we are studying any one variable, but only one at a time
• We are NOT trying to study the effect of an input variable on an output—that is 2-Variable testing
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Why Do We Care About “1-Variable” Testing?
• STABILITY: Is there special cause or common cause variation in our data impacting process stability?
• SHAPE: Is the distribution shape as we might expect, or are some problems, such as mixed processes, evidenced in the data?
• SPREAD & CENTER: Is the process failing to perform (meet expectations and requirements) with respect to target, variation, or both?
• We did some of these things in Define and Measure, but now we are doing it with validated data
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
• Hypothesis tests are designed to evaluate only one characteristic at a time
• E.g., “centering” tests tell us nothing about difference in spread
• In order to fully evaluate a variable’s characteristics, we have to employ several tests
Stability, Shape, Centering & Spread
9/16/2015
95
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
1-Variable: Continuous Data
Stability
• Run Chart
• Control Chart
Shape
• Histogram
• Probability plot
• Normality: Stat-Basic Stat-Graphical Summary
Spread
• If Std. = Std. Deviation or variance, run 1 variance test
• If Std. = Range, convert to Std. Deviation first, then run 1 variance test
• If distribution is non-normal, run 1 variance test and use ADJUSTED Results
Centering
• 1-sample t test – normal data (compare sample mean to customer request)
• 1-sample z test if population std. deviation is known (Rare)
• If non-normal, run 1-sample sign test
• If non-normal and distribution is relatively symmetrical, run 1-ample Wilcoxon (Rare)
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Stability: Run Charts
• Run charts detect for trends, clustering, mixtures, or oscillations in the process
• While control charts are a more powerful statistical tool, run charts have one advantage over control charts in that they can be used with non-normal data
• Always check for normality before using any type of control chart
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Run Chart Hypotheses
• H0 (Null hypothesis):
– Data is random, special causes not present
• HA (Alternate hypothesis):
– Special causes (mixtures, clusters, trends, or
oscillation) are present
• Data must be entered in time order!
9/16/2015
96
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
What Do I Do If My Process Is Not Stable?
• Instability is not necessarily a reason to stop, but it does cause us to re-evaluate the reliability of things like the baseline capability estimates.
• If instability is the result of special causes what should you do?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Shape: Normality
• Anderson Darling Test for Normality • Stat- Basic Stat- Graphical Summary
• Need at least 50 data points to get shape info!– Ho:– Ha:
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Testing for Spread
• Use 1 variance test to evaluate your data spread versus a std. (customer specification or Champion expectation)– If you are given std. deviation as a std., select
“enter std. deviation” in top drop-down box in 1-variance test
– If you are given variance as a std., select “enter variance” in top drop-down box in 1-variance test
– If you are given range as a std. to compare your data to, first convert range to std. deviation (see next slide), then follow procedure above for entering the std. deviation as a std. to compare to.
9/16/2015
97
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
1-Sample t-Test: Hypothesis testing for Means
• Hypothesis testing confirms (or not) a statement regarding centering or performance to target for our process
• This procedure is based upon the t-distribution, which is derived from a normal distribution with unknown s
• Use 1-Sample t to perform a hypothesis test of the mean when the population standard deviation, σ, is unknown– Ho:
– Ha:
• If the population std. deviation (σ) is known, use the 1-sample z test
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
1-Variable: Discrete Data
Stability
• Use the attribute control chart table; Alpha risk is Known; Beta risk is Unknown
Centering
• For defectives data, run 1 proportion test (pass/fail)
• For defect data, run 1-sample Poisson rate test (defect rates)
• For Nominal data, run Chi-Squared Goodness of fit (not powerful)
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
1-Variable Hypothesis Testing (Discrete Data)
• Testing behavior over time – Discrete/Attribute (control) charts
• Tests for Nominal data– Goodness of fit testing
– Chi-square tests
• Tests for Ordinal data– 1-sample test for
proportions
– 1-sample Poisson rate
– 1-sample sign test
– 1-sample Wilcoxon test
– Kolmogorov-Smirnov
9/16/2015
98
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
1-Sample Sign Test for Medians• One-sample t-tests look at means, while the
sign tests look at medians• H0: • Ha:
• If your distribution is relatively symmetrical, use 1-sample Wilcoxon test as it is more statistically powerful (can detect smaller differences) than the 1-sample sign test
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Analyzing Discrete Data
• Remember that discrete data is “information poor” relative to continuous data.– Using discrete data requires much larger
sample sizes (often times 100 or more data points) to achieve statistical power comparable to continuous data tests
• Often we overlook the importance of validating measuring system reliability for discrete data
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The Chi-Square Goodness of Fit Test
• We use Chi-Square test of independence to evaluate whether observed frequencies agree with expected frequencies
• Create table in minitab and enter the observed vs. expected data
• Run the Test (Minitab Demonstration)
9/16/2015
99
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov295
1-Sample Poisson Rate
• Use if your project metric is discrete rate:– 12 defects / hr, 17 correct items / invoice,
• Check your data performance versus a std. by using the 1-sample Poisson rate test– STAT>BASIC STAT>1-sample Poisson rate
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: 1 Variable
• 1 Variable Testing: Standard of completing tickets is a half-day or 240 minutes
• Answer Questions• *Minitab Demonstration*
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Box PlotsDMAIC – Understand Metrics
9/16/2015
100
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Box Plots
• Box Plots: (box-and-whisker plots) They are particularly useful for showing the distributional characteristics of data such as center and spread.
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Box Plots
• A box plot consists of a box, whiskers, and outliers.
• A line is drawn across the box at the median. The bottom of the box is at the first quartile (Q1), and the top is at the third quartile (Q3) value. The whiskers are the lines that extend from the top and bottom of the box to the adjacent values.
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Class Activity: Boxplot
• Class Discussion: What x’s could we utilize boxplots in the help desk process?
• Minitab Demonstration: Boxplot demonstration from the above discussion
9/16/2015
101
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
2-Variable: Continuous
DMAIC – Determine Significant X’s
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Road Map: Compare 2 samples with each other
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Continuous Y, Discrete X with 2-levels
• In this section, we will learn how to conduct hypothesis tests for a continuous Y variable and a discrete X having only two levels.– 2-Sample t-test for comparing means from two
independent groups– Mann-Whitney nonparametric test for two medians
• Learn how to use “residuals” to verify required assumptions
9/16/2015
102
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Continuous Y, Discrete X with 2-levels
• Continuous Y data, and Discrete X data• The X variable has exactly two levels (settings),
producing two groups of Y data• Uses one of the several forms of t-tests• Assumptions:
– Stable process in both groups– Equal variance for both groups– Normal distribution within each group
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The General Approach
• Using the hypothesis testing roadmap for two-variable testing, and the characteristics of the variables, identify the appropriate test category
• Perform appropriate graphical examinations as needed
• Conduct the hypothesis test(s)• Verify that any required assumptions regarding
shape, center, spread, and stability are satisfied
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The General Approach
• At any of these steps along the way, we could (and probably should) apply graphical tools to aid in the analysis
• Don’t worry. In most cases you will be using many of the same graphical tools that you’ve already seen
9/16/2015
103
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
“Residuals” Will Set Us Free!• All hypothesis tests have at their core some
mathematical basis, or model• When we establish a null hypothesis, we make those
models specific to our problem• In a perfect world, these models would fit our data
perfectly, however in our real, imperfect world there will be some data that doesn’t fit the model perfectly
• We can use this imperfect (residual) data to evaluate how well our data fits the assumptions in the mathematical model. If the residuals show acceptable behavior, then the assumptions are considered valid
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
“Residuals” Will Set Us Free!
• The nice part about using residuals to check the assumptions is that Minitab will calculate them for you automatically and…
• …they can be calculated for all of the groups in one operation!!
• This is almost always accomplished in Minitab by just turning on an option switch to have Minitab calculate and save the residuals
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
“Residuals” Will Set Us Free!
• A good fitting model will have residuals that behave the same way regardless of the level of the X variable. In other words, we don’t have to check assumptions for each separate group.
– All groups will have the same shape (Normal)– All groups will have the same variance– All groups will be stable (no patterns or special c auses)– No bias in the results
• If any of the above are shown false, then at least one of the required assumptions has failed, and the test results would then be suspect
9/16/2015
104
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
As an example, suppose a supervisor wants to know if two operators process calls faster or differently
The Testing Roadmap First Asks for Data Types
• What’s the Y ? ________ Type of Data? ____________
• What’s the X ? ________ Type of Data ? ____________
• How many “levels” does X have? ________________________
• What type of tool would you use ? ________________________
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The First Step is to Prepare the Data For Analysis
• Minitab’s procedures are designed to have all of the variables arranged in columns– A single column for the Y variables– One column for each of the X variables
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Types of Hypotheses for t-Test
• Two-tailed test– Used when the direction of the differences between two
means cannot be stated up front, or does not matter -set up as an equality
• One-tailed test– Used when the direction of the differences between two
means can be stated up front - set up as an inequality
9/16/2015
105
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Ho: ( 0) Ha: ( 0)δ δ= ≠
43210-1-2-3-4
0.4
0.3
0.2
0.1
0.0
Output
Two-Tailed Test
T=1.96T=-1.96
Example: = 1δ µ µ− 2
For α = .05
α 2 025=. α 2 025=.
Two-Tailed Test
• Hypothesis Tests usually can be stated in terms
of
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Right-tailed Test
43210-1-2-3-4
0.4
0.3
0.2
0.1
0.0
Output
T=-1.64
α =.05
Example: = 1δ µ µ− 2
Ho: ( 0) Ha: ( < 0)δ δ≥
43210-1-2-3-4
0.4
0.3
0.2
0.1
0.0
Output
T=1.64
α =.05
Ho: ( 0) Ha: ( > 0)δ δ≤
Left-tailed Test
For α = .05
One-tailed Tests
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The t-Test is Built on Certain Assumptions
• Remember that the t-test results are only completely valid if the underlying assumptions are true– Equal Variance – Normality– Stable Process
• We will check these assumptions using “residuals”, Residuals will be covered in the ANOVA section
9/16/2015
106
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The Mann-Whitney Test
• The Mann-Whitney test is a nonparametric equivalent to the t-test
• It does not require the data to be normally distributed--it works for any shape
• However, it does require that both groups have approximately the same shape
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Discussion*: 2-Variable
• Based on the Help Desk Process: What Xs require 2-Variable Testing?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
2-Variable: DiscreteDMAIC – Determine Significant X’s
9/16/2015
107
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Chi-Square Test
• If the response variable (Y) is a count– Even if it is expressed in the form of a rate or
proportion, we can usually get the count that was used to calculate the rate or proportion
– The Chi-Square test uses observed counts in the various combinations of Y and X to determine if there is a significant relationship between X and Y
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The Personnel Department wants to see if there is a link between age (old and young) and whether that person gets hired
Discrete Y and Discrete X
• What’s the Y ? ________ Type of Data ? ____________
• What’s the X ? ________ Type of Data ? ____________
• What type of tool would you use ?
________________________
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Frequency Expected f
Frequency Observed f
:where
2
e
o
)(
=
=
2 ∑ −=e
eo
fffχ
What is the Chi-Square Distribution?
• Chi-Square is based on expected and observed frequencies (counts) for all combinations of the X and Y variables
9/16/2015
108
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Hired Not Hired
Old 30
45
150
230Young
Total
Total 75 380 455
275
180
How do you make the decision here ?
The Data
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Ho: Data are Independent (Not Related)
Ha: Data are Dependent (Related)
If the P Value is <.05 , then reject Ho
The Hypothesis
• With the Chi-Square Test for Independence,
statisticians assume most variables in life are
independent, therefore:
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The Theory
• Let’s walk through our example….• Assume we wanted to determine if age and
hiring practices are dependent or independent
• Therefore our hypotheses are stated as follows...– Ho:– Ha:
Age and Hiring Practices are independent
Age and Hiring Practices are dependent
9/16/2015
109
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Hired Not Hired
Old 30
45
150
230Young
Total
Total 75 380 455
275
180
How do you make the decision here ?
The Data
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Analyzing the Data in Minitab
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Chi-Square Test
Expected counts are printed below observed counts
Hired Not Hire Total1 30 150 180
29.67 150.33
2 45 230 27545.33 229.67
Total 75 380 455
Chi-Sq = 0.004 + 0.001 +0.002 + 0.000 = 0.007
DF = 1, P-Value = 0.932
What Decision Would You Make?
A P-Value !
Analyzing the Data in Minitab
• Note: The observed and expected counts are the same values you calculated a moment ago
9/16/2015
110
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Ho: Data are Independent (Not Related)
Ha: Data are Dependent (Related)
If the P Value is <.05 , then reject Ho
The Hypothesis
• With the Chi-Square Test for Independence, statisticians assume most variables in life are independent, therefore:
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Chi-Square Comments
• Chi-Square is the least “insightful” and usually one of the more “difficult to analyze” tools we learned this week– But that is what happens when we deal with
discrete data!• You must have at least FIVE data points for
the Chi-Square Test to work or Minitab will crash
• Your data should have been gathered to assure randomness
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
ANOVA: ContinuousDMAIC – Determine Significant X’s
9/16/2015
111
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Road Map: Compare 2 samples with each other
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
A Continuous Y, and a Discrete X but >2 levels
• There is a great deal of similarity between the ANOVA methods and the t-test methods– In fact, the t-test is really a special case of general
ANOVA– For the equal variance case, the t-test and the ANOVA
will produce exactly the same results
• Just as with the t-tests, ANOVA is built upon some foundation assumptions that must be verified
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
The Concept of a Single Factor Experiment
• When we have multiple levels of a single input variable (also known as a “factor”), we are really analyzing the results of what is usually called a “Single Factor Experiment”– The t-test is a single factor experiment because there was 1-Y
variable, and 1-X variable with two levels– Note that in a single factor study we are not limited to only two
levels of the input variable
• We are usually trying to identify if there is any difference between the different levels of the factor under investigation– Example: Looking at 5 different regions to determine if they are
consistent
9/16/2015
112
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Single Factor Experiments
• Here is an example: – Ho: All the students will score the same on
the test – Ha: At least one student will score
differently on the test
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Statistical Assumptions
• The data comes from a stable process
• Each group is independent from each of the others, and the data within a group is approximately normal in distribution– Residuals are independently and normally
distributed with a mean of 0 for each group
• The variances of the result (Y variable) are equal across all levels of the input variable– The residuals will have a constant variance across
groups
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
One Way ANOVA
• After running the ANOVA, we still have to confirm the assumptions (check the residuals)
• The good news is that this time Minitab will produce the residuals for us…– We must remember to set the “Store Residuals”
switch if we want Minitab to produce the residuals for us
9/16/2015
113
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
One Way ANOVA
• By selecting the “Store Residuals” switch, Minitab will calculate all of the residual values and put them in a new column in the worksheet
• While you may not know the reason yet, go ahead and also turn on the “Store Fits” switch as well– We’ll use this a little later
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Residuals Analysis
• Letting Minitab calculate the residuals saves some work
• Still need to verify the behavior of the residuals:– Equal Variance– Normally Distributed– Stable Process
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Residuals Analysis
• If the residuals fail then we need to use a nonparametric test
• The Kruskal-Wallis and the Mood’s Median tests are nonparametric equivalents to the one-way ANOVA– Mood’s Median is used for data that has outliers– Kruskal-Wallis is more powerful (no outliers in data
set)
• Again, neither nonparametric test is more powerful than the ANOVA
9/16/2015
114
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: ANOVA
• Based on the Help Desk Process what Xsrequire ANOVA?
• Complete the Hypothesis Form for two scenarios
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Minitab Demonstration*: ANOVA
• Demonstrate the Hypothesis Tests from the previous Class Activity
• Interpret your results - what X’s are/are not significant?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Simple RegressionDMAIC – Determine Significant X’s
9/16/2015
115
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Regression Definition
• Regression analysis is a statistical technique used to investigate and model the relationship between variables
• Simple Linear Regression relates one continuous Y with one continuous X
• Multiple Linear Regression relates one continuous Y with more than one continuous X (Covered in Week Four)
• Model adequacy is checked by reviewing the quality of the fit
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
50403020
500
400
300
200
100
0
Exp. Level
Res
p_T
ime
S = 28.6842 R-Sq = 93.3 % R-Sq(adj) = 93.3 %
Resp_Time = -278.749 + 12.9680 Exp. Level
Regression Plot
Regression - Fitted Line Plot
• From the Graphic, we can see that as experience level increases, so does response time
• We say that there is a “positive” relationship here
• We also see a linear equation and an R-Sq value
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Fitted Line Plot - Session Window Output
• The Minitab output recaps the regression equation and R-Sqstatistics
• It also provides an ANOVA table
• In fact, ANOVA is really just a special case of general regression
Regression Analysis: Resp_Time versus Exp. Level
The regression equation is Resp_Time = -278.749 + 12.9680 Exp. Level
S = 28.6842 R-Sq = 93.3 % R-Sq(adj) = 93.3 %
Analysis of Variance
Source DF SS MS F PRegression 1 1619241 1619241 1968.00 0.000Error 141 116012 823 Total 142 1735253
What do we conclude?
9/16/2015
116
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Influence of “X” Variables
• If the outlier is a bad value, then the model estimates are wrong and the error is inflated
• However, if the outlier is a real, process value, it should not be removed. It is a useful piece of data on the process.
• Evaluate the model with and without the point to determine its effect
X= Predictor
Y =
R
espo
nse
What is the effect on the regression coefficient
with the observed outlier?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
0 1 2 4 5X= # of Storks
Y =Town
Population
15
10
5
“Nonsense” Relationships or Wrong Conclusions
• Data on a city showed that as population density of storks increased, so did the town’s population. Did storks influence the population ?
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Regression
• What are the two Continuous data sets in the Help Desk Process?
• Which explains the other?
*Minitab Demonstration*: Conduct Regression for Help Desk Process and Interpret Results
9/16/2015
117
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Update Project Charter
• Review the Project Charter and make updates if appropriate.
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
IMPROVE
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Improve
• Improve Purpose: To develop and quantify potential solutions, improve/optimize the process, evaluate and select final solution and implement the pilot.
1. Establish level for X’s2. Develop Solutions
3. Pilot and Implement
9/16/2015
118
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Improve Tools (Additional Information)
• Brainstorming• Affinity• Prioritization of
Ideas• Lean Improvement
Solutions• Mistake Proofing
• 5S• Kanban• Optimal Settings
for X’s• Implementation
Plans• Change
Management• Innovation
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Improve
• After conducting all of our test, reviewing the current state process map, and VOC what are your recommendations to create a better process?
• Prioritize the improvement ideas
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Operational Definitions
• One of the main issues is lack of prioritization of type of ticket
• The team has decided to implement a priority ranking for the ticket types
• Please create an operational definition for the four priority types and place the main ticket types into each priority
9/16/2015
119
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
CONTROL
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Control
• Control Purpose: Implement final solution, maintain process improvements, ensure new process problems are identified and quickly corrected, disseminate lessons learned, identify areas for replication and standardization.
1. Evaluate Process Performance (Results)2. Develop Control Plan3. Transition to Project Owner
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
Control Tools
• Action Registries• 30-60-90 Follow-up
Meetings• Control Plans
• Transition Plans• Capability Analysis
• Control Charts
9/16/2015
120
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Control
New Data: January and February • What information would you want to
know?
• *Class Activity-Questions* : Review the new data. Be prepared to report out initial findings. Be prepared to report out what tests you would like to run.
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Minitab Demonstration*: Control
• Run through identified tests
• *Class Discussion*: Interpret test results
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Update Project Charter
• Review the Project Charter and make updates if appropriate.
9/16/2015
121
SIMPLER. FASTER. BETTER. LESS COSTLY. lean.ohio.gov
*Class Activity*: Report Out
• Complete your report out• Make team assignments
• Report out time: 15 minutes maximum