36
Statistical Methods College of Science Department of Statistics Statistical Methods Department of Statistics Dr. Rick Edgeman, Professor & Chair and Six Sigma Black Belt Tel. +1 208-885-4410 Fax. +1 208-885-7959 Email: [email protected]

S tatistical M ethods

  • Upload
    tab

  • View
    32

  • Download
    0

Embed Size (px)

DESCRIPTION

D epartment of S tatistics. S tatistical M ethods. D r. R ick Ed geman, P rofessor & C hair and S ix S igma B lack B elt Tel. +1 208-885-4410Fax. +1 208-885-7959Email: [email protected]. The Scientific Method. Noninformative Event. Informative Event. Little or Nothing Learned. - PowerPoint PPT Presentation

Citation preview

Page 1: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Statistical Methods

Department of

Statistics

Dr. Rick Edgeman, Professor & Chair and Six Sigma Black BeltTel. +1 208-885-4410 Fax. +1 208-885-7959 Email: [email protected]

Page 2: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

The Scientific Method

No Observeror Uninformed

Observer

InformedObserver

Noninformative Event Informative Event

Scientific Method

of Investigation

Nothing Learned

Little orNothing Learned

Little orNothing Learned

Discovery!!

Page 3: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Conjecture

DesignAnalysis

Experiment-Data

Planned

Change

The Design of Experiments (DOE) Approach

Page 4: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

The Hypothesis Testing Approach

Conjectures (Hypotheses)

Information & Risk

Requirements

Evaluation

(Test M

ethod)D

ecisionC

riteria

Gather & Evaluate

Facts

Mea

ning

& A

ction

(s)

Infor

med

Decisi

on

Zon

e of

Bel

ief

ConsequencesA B… or …

Page 5: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Motivation for Hypothesis Testing• The intent of hypothesis testing is formally examine two

opposing conjectures (hypotheses), H0 and HA.

• These two hypotheses are mutually exclusive and exhaustive so that one is true to the exclusion of the other.

• We accumulate evidence - collect and analyze sample information - for the purpose of determining which of the two hypotheses is true and which of the two hypotheses is false.

• Beyond the issue of truth, addressed statistically, is the issue of justice. Justice is beyond the scope of statistical investigation.

Page 6: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics The American Trial System In Truth, the Defendant is: H0: Innocent HA: Guilty

Correct Decision Incorrect Decision

Innocent Individual Guilty Individual Goes Free Goes Free

Incorrect Decision Correct Decision

Innocent Individual Guilty Individual Is Disciplined Is Disciplined

Innocent

Guilty

Ver

dic

t

Page 7: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics True, But Unknown State of the World H0 is True HA is True

Ho is True

Decision

HA is True

Correct Decision Incorrect Decision Type II Error Probability =

Incorrect Decision Correct Decision Type I Error Probability =

Page 8: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics Hypothesis Testing & The American Justice System

• State the Opposing Conjectures, H0 and HA.

• Determine the amount of evidence required, n, and the risk of committing a “type I error”,

• What sort of evaluation of the evidence is required and what is the justification for this? (type of test)

• What are the conditions which proclaim guilt and those which proclaim innocence? (Decision Rule)

• Gather & evaluate the evidence.

• What is the verdict? (H0 or HA?)

• Determine a “Zone of Belief” - Confidence Interval.

• What is appropriate justice? --- Conclusions

Page 9: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Study

DoAct

Plan

Planned

Change

The Plan-Do-Study-Act (PDSA) Approach

Note: A.K.A. theShewhart Cycle orDeming Wheel

Page 10: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Study

DoAct

Plan

Planned

Change

A Modified Plan-Do-Study-Act (PDSA) Approach

Study(Baseline Description)

Standardize

Page 11: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

• Baseline Evaluation: Capture / describe initial process performance.• Plan: What could be the most important accomplishment of this team? What

changes might be desired? What data are available? Are new observations needed? If yes, plan a change or test. Decide how you will use the observations.

• Do: Search for data on hand that could answer the question put forth in the P stage. Or, carry out the change or test decided upon, preferable on a small scale. This is often a Reduced Implementation.

• Study the effects of the change or test.• Standardize the approach with an eye toward portability of solution.• Act: What actions should be taken? This is often more extensive

implementation.• Hold the Gain: Prove the gain to be sustainable. This may be concurrent with

the next planning stage of this repetitive cycle.• Iterate as needed … this is a cycle!

Modified PDSA Cycle Description

Page 12: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Define-Measure-Analyze-Improve-Control (DMAIC) Approach

Define

Control

Improve Analyze

Measure

Planned Change

Six

Sigma

Innovation

Page 13: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Are TQM & Six Sigma the Same? Are Six Sigma Efforts Always Successful?

Page 14: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

The Six Sigma Strategy Six Sigma Strategy Affects Six Areas

Fundamental to Improving a Company’s Value:1. Process Improvement2. Product & Service Improvement3. Investor Relations4. Design Methodology5. Supplier Improvement6. Training & Recruitment

Page 15: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

SIPOC Model

Suppliers Customers

Inputs OutputsProcess

Steps

Inform Loop

Page 16: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

COPIS Model

Customers Suppliers

Outputs InputsProcess

Steps

SIPOC from a Six Sigma Perspective: From the Six Sigma Perspective, the model is a “COPIS” one in the sense that Six

Sigma projects are customer-driven, begin with the customer, and are pushed back through the value chain to the supplier.

Page 17: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Six Sigma’s DMAIC Innovation & Improvement Algorithm

Define Control

Measure ImproveAnalyze

Voice Of the Customer

Institutionalization

Page 18: S tatistical  M ethods

Statistical Methods

College of Science Department of

StatisticsStage Objective Phase Detail

Identification Identify Key Business Issues

Recognize

DefineDefine the Problem and Customer Requirements.

Characterization Understand Current Performance Levels Measure

Measure Defect Rates & Document the Process in its Current Form.

AnalyzeAnalyze Process Data & Determine the Capability of the Process.

Optimization Achieve Breakthrough Improvement

ImproveImprove the Process and Remove Defect Causes.

ControlControl Process Performance and Ensure that Defects do not Recur.

Institutionalization Transform how Day-to-Day Business is Done

Standardize

Integrate

Bla

ck B

elt

Pro

jects

Page 19: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Six Sigma from the GE Perspective:

Six Sigma is a highly disciplined process that helps a company focus on developing anddelivering near-perfect products and services. Why “sigma”? The word is a

a statistical term that measures how far a given process deviates from perfection.

The central idea behind Six Sigma is that if you can measure how many “defects” youhave in a process, you can systematically determine how to eliminate those

and approach “zero defects”.

Six Sigma has changed the DNA at GE – it is the way that GE works – in everything that GE does and in every product GE designs.

“What is Six Sigma? The Roadmap to Customer Improvement”www.ge.com/sixsigma/makingcustomers.html

Page 20: S tatistical  M ethods

Statistical Methods

College of Science Department of

StatisticsSix Sigma Quality

Definition •Quality is a state in which value entitlement is realized for the customer and provider in every aspect of the business relationship.

•Business Quality is highest when the costs are at the absolute lowest for both the producer & consumer.

•Six Sigma provides maximum value to companies in the forms of increased profits and maximum value to consumers with high-quality products and services at the lowest possible cost.

Page 21: S tatistical  M ethods

Statistical Methods

College of Science Department of

StatisticsSix Sigma & the Cost of Poor

Quality The cost to deliver a quality product can account for as much as 40% of the sales price.

For example, a laser jet printer purchased for $800 may have cost the manufacturer $320 in rework just to make sure that you took home an average-quality product.

For a company whose annual revenues are $100 million and whose operating income is $10 million, the cost of quality is roughly 25% of the operating revenue, or $25 million.

If this same company could reduce its cost of achieving quality by 20%, it would increase its operating revenue by $5 million – or 50% of the current operating income.

Cost of Quality and DPMO

DPMO Cost of Quality

2 308,537 Not Applicable 3 66,807 25%-40% of sales4 6,210 15%-25% of sales5 233 5%-15% of sales6 3.4 < 1% of sales

Each sigma shift provides a 10% net income improvement..

Page 22: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Structured Problem-Solving With DMAIC:The Heartbeat of Six Sigma

Page 23: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Six Sigma Projects Begin with aDetailed Assessment of Customer Needs

Define:

A. Identify project CTQs: what does the customer think is essential?

B. The Team Charter represents the business case for the project.

C. Define and build a process map that relates measurableinternal processes to customer needs.

These will now be addressed in greater detail

Page 24: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Define: A. Identify project CTQs: what does the customer

think is essential?

Voice Of the Customer (VOC)That which is critical to the quality of the process according to your

customer.

VOC tools:Surveys

Focus Groups Interviews

Customer Complaints

Page 25: S tatistical  M ethods

Statistical Methods

College of Science Department of

StatisticsAdvantages: Lower cost approach Phone response rate 70-

90% Mail surveys require least

amount of trained resources for execution

Can produce faster results

Disadvantages: Mail surveys can get incomplete

results, skipped questions, unclear understanding

Mail surveys 20-30% response rate

Phone surveys: interviewer has influential role, can lead interviewee, producing undesirable results

Advantages: Group interaction generates

information More in-depth responses Excellent for getting CTQ

definitions Can cover more complex

questions or qualitative data

Disadvantages: Learning’s only apply to those

asked, difficult to generalize Data collected typically

qualitative vs. quantitative Can generate too much

anecdotal information

Focus GroupsFocus Groups

SurveysSurveys

Page 26: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Advantages: Specific feedback Provides opportunity to

respond appropriately to dissatisfied customer

Disadvantages: Probably not adequate sample

size May lead to changing process

inappropriately based on 1-2 data points

Advantages: Can tackle complex

questions and a wide range of information

Allows use of visual aids Good choice when

people won’t respond willingly and/or accurately by phone/mail

Disadvantages: Long cycle time to complete Requires trained,

experienced interviewers

Customer ComplaintsCustomer Complaints

InterviewsInterviews

Page 27: S tatistical  M ethods

Statistical Methods

College of Science Department of

StatisticsSurvey Development

Information What do I need to know when this study is complete? What is my budget? What information will the survey provide that cannot

be obtained elsewhere? How much time do I have to complete the study? Who will be surveyed and how do I reach these

people?

Page 28: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Survey Development Steps

• Review survey objectives.• Determine appropriate sample.• Identify specific areas of desired information.• Write draft questions and determine measurement scales.• Determine coding requirements.• Design the survey.• Pilot the survey–both the individual questions as well as the

total survey against the objectives.• Revise and finalize.

Creation of Electronic Surveys: www.zoomerang.com

Page 29: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Define:A. Identify project CTQs: what does the customer think is essential?

Who is the customer and what do they want? This may be derived from:Business Goals; Complaint Information; Customer Surveys or Focus Groups;

Benchmarking Data; Executive-Level Discussions; or Job-Specific Discussions.

We need a “Process / Product Drill-Down Tree”Y = f(X1, X2, …)

“Big Y” is a function of X1, X2, … where the X’s are internal process characteristicsor ‘CTQs’ that can be controlled. CTQs represent customer desired outcomes.

Drill Down Trees Integrate Customer CTQs and Business Strategy.

In this drill down tree the “Big Y” is decomposed into “little y’s” that are subprocesses of Y.This “drill down” continues through DEFINE and MEASURE. The X’s are part of ANALYZE.

Page 30: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

SMART Problem & Goal Statements Are:

Specific

Measurable

Attainable

Relevant

Time-Bound

Page 31: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Project Scope On what process will the team focus on? What are the boundaries of the process we are to improve? Start

point? Stop point? What resources are available to the team? What (if anything) is out-of-bounds for the team? Under what (if any) constraints must the team work? What is the time commitment expected of team members? What are the advantages to each team member for the time

commitment?

Page 32: S tatistical  M ethods

Statistical Methods

College of Science Department of

StatisticsEight Steps for

Establishing

Project

Boundaries

1. Identify the customer– Who receives the process output?

(May be an internal or external customer)2. Define customer’s expectations and needs

– Ask the customer– Think like the customer– Rank or prioritize the expectations

3. Clearly specify your deliverables tied to those expectations– What are the process outputs? (Tangible and intangible deliverables)– Rank or prioritize the deliverables– Rank your confidence in meeting each deliverable

4. Identify CTQ’s for those deliverables– What are the specific, measurable attributes that are most critical in the

deliverables?– Select those attributes that have the greatest impact on customer

satisfaction.

Page 33: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

5. Map your process– Map the process at it works today (as is).– Map the informal processes, even if there is no formal, uniform process in

use.6. Determine where in the process the CTQ’s can be most seriously affected

– Use a detailed flowchart– Estimate which steps contain the most variability

7. Evaluate which CTQ’s have the greatest opportunity for improvement– Consider available resources– Compare variation in the processes with the various CTQ’s– Emphasize process steps which are under the control of the team conducting

the project8. Define the project to improve the CTQ’s you have selected

– Define the defect to be attacked

Eight Steps for

Establishing

Project Boundaries

Page 34: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Measure: Define Performance Standards: Numbers & Units

Translate customer needs into clearly defined measurable traits.

OPERATIONAL DEFINITION: This is a precise description that removes any ambiguity about a process and provides a clear way to measure that process. An operational definitionis a key step towards getting a value for the CTQ that is being measured.

TARGET PERFORMANCE: Where a process or product characteristic is “aimed”. If therewere no variation in the product / process then this is the value that would always occur.

SPECIFICATION LIMIT: The amount of variation that the customer is willing to tolerate in a process or product. This is usually shown by the “upper” and “lower” boundary which, ifexceeded, will cause the customer to reject the process or product.

DEFECT DEFINITION: Any process or product characteristic that deviates outside ofspecification limits.

Page 35: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics

Measure: Establish Data Collection Plan, Validate the Measurement System, and Collect Data.

A Good Data Collection Plan:

a. Provides clearly documented strategy for collecting reliable data;b. Gives all team members a common reference;c. Helps to ensure that resources are used effectively to collect only critical data. The cost of

obtaining new data should be weighed vs. its benefit. There may be historical dataavailable.

We refer to “actual process variation” and measure “actual output”:a. what is the measurement process used? b. describe that procedure c. what is the precision of the system? d. how was precision determinede. what does the gage supplier state about: f. Do we have results of either a:

* Accuracy * Precision * Resolution * Test-Retest Study? * Gage R&R Study?

Page 36: S tatistical  M ethods

Statistical Methods

College of Science Department of

Statistics Establish Data Collection Plan, Validate the Measurement System, and Collect Data.

Note that our measurement process may itself have variation.

a. Gage Variability:

Precision Accuracy Both

b. Operator Variability: Differences between operators related to measurement. c. Other Variability: Many possible sources. Repeatability: Assess effects within ONE unit of your measurement system, e.g., the variation in the measurements of ONE device. Reproducibility: Assesses the effects across the measurement process, e.g., is there variation between different operators. Resolution: The incremental aspect of the measurement device.

Measure