Sigma Six Study Guide A Attribute data is discrete Appraisal- ex: cost of calibration and first trial test hardware Affinity diagram: used in define
Anova test for comparing 3 or more means are that the populations are independent, Normal and that they ha ve homogenous (equal) variances
Analyze Phase: identify the root cause of process variability B Binomial- (n/r(n-r))*P^rQ^n-r
Brainstorming and prior operational knowledge are used during the design phase of the experiment to gather process information, help determine the objectives of the study and to identify potential input factors.
Breakthrough: describes the desired change in performance level
Blocking is used to account for the variation due to known sources such as people, machines, etc. that are not of direct interest to the study, in order to minimize their impact on the total variation. Balanced Scorecard: Corporate, Sales, Internal, Infrastructure baseline performance level: Provide a starting reference point to gage improvement Box Plots are used to investigate the effect of discrete inputs on continuous outputs. Several levels of inputs can be compared side-by-side, making this an easy-to-understand tool for use with a team
C Cyclical variation can be displayed as variation among consecutive pieces, variation among groups of pieces, variation among consecutive batches Conveyance is moving a single part excess times
Continuous flow is to reduce the batch (or lot) size to one
Change Agent: overcome fear of unkown Check Sheet: used for collecting data to study the symptom of a problem Cp tells you=what is possible if your process is perfectly centered, process potential, relative ratio of the process variation to the specifications range, the best you process can be o Cp is less or equal that Pp o Process becomes centered the Cpk or Ppk increases o Cp of 3 means very little product (1%) is out of specification o Cp and Cpk are for short term data sets o Sigma level increases-Ppk and Cpk increase o The process data must be normally distributed in order for any capability index (Cp, Cpk, Pp, or Ppk) to
correctly reflect capability. Transformation may be accomplished using any of several different algorithms to achieve a fit to the normal distribution.
Contingency Tables- compare more than two sample proportions with each other Cpk equals 1 when +- 3 sigma and normal center distribution
Cp=(USL-LSL)/6 = (5-3)/6*0.25 = 1.33. Cpk=min((USL-process mean)/3 , (process mean-LSL)/ 3) = min((54.2)/3*0.25 , (4.2-3)/3*0.25) = min(1.07,1.6) = 1.07.
Coefficient of variance: standard deviation/mean
complete fold-over adds a second fraction of runs to the design with the signs of all factors reversed, thus breaking the aliases between all main effects and two-factor interactions. Note: A single factor fold-over only breaks the aliases on the specific factor chosen for the fold-over.
Criteria to selected a project: B/(I +C)>1
Control Plan is used to keep track of the status of all significant characteristics, after they have been identified. The CHECK Process is used to evaluate process conformance to specifications. It can ensure that input factors are set at the specified levels, but cannot verify that these levels are optimal
Control Charts: determines process capability CTQC: The Overall Importance Ranking prioritizes Causes (Inputs) in terms of their leverage on the most important
Effects
Cycle Time: used for evaluating the tangible effects of a six sigma project Coding- method that changes data without significantly reducing accuracy A confidence interval provides a range of values which will contain the true value of the parameter (the Population proportion/mean/variance etc.) based on the confidence level selected. For example, a 95% confidence interval gives a range of values which will contain the true value of the parameter (the Population proportion/mean/variance etc.) for 95 out of every 100 samples. Withinappraiser (Repeatability) Condition: Two or more trials per appraiser. Number of agreements (a) = the number of units which have the same rating across all trials for a given appraiser.Between-appraiser (Reproducibility)Condition: Two or more appraisers in the study. Number of agreements (a) = the number of units which have the same rating across all trials for all appraisers.When there is a standard/reference available in addition to the above conditions,Each appraiser vs. Standard Number of agreements (a) = the number of units which have the same rating across all trials for a given appraiser and the standard. All appraisers vs. Standard Number of agreements (a) = the number of units which have the same rating across all trials for all appraisers and the standard. Let n = the total number of units inspected. Then for each of the above the Percent agreement is given by P = (a/n)*100
D Data types- variable and attribute Defect: A product/process that does not meet customer requirements for use.
Diamond is a decision point Design of Experiments: to improve/characterize the process or product; bringing the process on target and minimizing variation might result in an in-control process, but that is not the purpose of DOE. Design for Manufacturability: Use the minimum number of parts possible.
E End user dissatisfaction- increase in the percentage of returned goods
error-proofing are: Warnings, Shutdown and Autocorrection.
External Set-up activities are defined as those which take place while the process is running
F Factorial Factor Designs- quarter fractional experiments-4 factors, Resolution 5 or higher desired, Functional Requirements is traceable to the voice of the customer Force Field Analysis: helpful in identifying the "drivers" and "restraints", used in initial stage of planning Flowchart- improve existing process by first gaining a clear understanding of the current process. FEMA- determine potential pitfalls and the actual defects that occur F-Test to be valid: populations normal, samples independent, samples must be randomly selected from the population
Fishbone Cause and Effect Charts are used to identify the potential sources of process variability, Fishbone Diagram and 5Why Analysis are tools designed to help identify the factors contributing to a problem. None of the other tools mentioned serve this specific purpose.
G Gage R&R- determines if measurements system has good repeatability and reproductively
gage calibration program: Systematic follow-up on scheduled calibrations, Track individual instruments with unique serial numbers, Provide trace-ability of standards to NIST
Gage variation- analysis designed to assess the statistical properties Hard Savings- regaining lost sales Hypothesis Testing is to make inference about the population House of Quality: customer requirements, product characteristics, competitive analysis Histogram: Display relative frequency by interval. Used primarily to show the centering, spread and shape of the
data by displaying the relative frequency of observations within categories(intervals).
I Identifying the functional requirement- used for evaluating and understanding design intent Internal Failure Cost- cost resulting from a nonconformance detected before a product or service is provided
Indicator/Predictor: warranty returns
J K Kaizen- Continuous incremental improvement
Kano Model: Over time, Excitement Quality characteristics tend to become Basic Quality requirements.
L Linearity is the difference in bias values across the process range Levene test of equal variance is used to compare the estimated population standard deviations from two or more samples with non-normal distributions
Logistic regression works for binary response data because it relaxes the other assumptions of OLS regression: Error terms are independent of each other
Level of significance- describe type 1 error M Median-used for skewed distribution
Mann-Whitney-Wilcoxon: response variable is not normally distributed, two independent samples; It is used in place of the independent samples t-test when the normality assumption of that parametric test is violated.
Multi Vari Charts: Used to analyze positional, cyclical, and temporal variations Moving Range: plots the mean of a set values of and recalculates the mean with each new value
measurement system: 30 % or less
N Nominal Group Technique: used when a team is generating and prioritizing a list of options that include highly controversial issues; least informative of the four measurement scales
Noise refers to uncontrollable sources (people, machines, shifts, etc.) which cause variability under normal conditions.
Norming phase: Competing loyalties, agendas, and responsibilities are reconciled Non-Parametric statistics methods when knowledge of the underlying distribution of the population is limited O One at a time experiment will miss interactions OEE: Availability=1-Losses in time due to equipment breakdowns=1-1.5/80=0.98125. Performance=1-Losses in time
due to minor line stoppages=1-2.5/80=0.96875. Quality=1-Losses due to defects =1-73/1000=0.927. So OEE=Availability * Performance * Quality Operational Definition: Provides a clear, unambiguous acceptance standard.
P Ppk purpose is to assess process performance, how well the process is actually running, capability relative to specification limits, and to see if your process is perfectly centered relative to the calculated Pp
Power of a test: P(Reject Ho/Ha is true)
Prevention- error proofing devices Pilot: reason include: reduce risk, Uncover unidentified issues, Allow for verification prior to
implementation, Reduce disruptions
Problem Statement achieves consensus and ownership of the process Pareto- prioritize processes
Process (Queue) = 16 hours x (2 Items/Minute x 60 Minutes) = 1920
Passive Study is multi-vari studies are a graphical tool that displays patterns of data performed as a passive study where the process is studied in its natural unchanged state Poison=(NP)^re^-np/r
Pooled t-test: used to compare 2 means when the population variances are unknown but assumed equal Project Charters: defines the mission
Poka-yoke-preventing controllable defects
p-value, based on the sampling distribution of the test statistic, is the probability of observing a value of the test statistic that is at least as extreme(high/low) as the calculated value, under the assumption that the null hypothesis is true.
Process cycle time: used to reduce process cycle time is to analyze current processes, reduce queue times, setting scheduling priorities
parametric tests: They make fairly strict assumptions about the distribution underlying the sampled data. make inferences about parameters, not sample statistics. They are used with continuous data but in most cases the data have to be normally distributed for the test to be valid and thus, they make fairly strict assumptions regarding the distribution underlying the data.
Process input: Loan Pricing, Credit Requirements, Loan Closing Cycle-time
p Chart would be correct because it measures fraction defective (percent defective), and can be used with either a variable or constant sample size. With a constant sample size, an X and Moving Range chart would also be appropriate (and a better choice due to greater sensitivity), but it was not a choice
Q Quality functional deployment- matrix tools are used, Translate voice of the customer data into
internal specifications
Quality Leadership: establish organizational goals to met or exceed customer needs, continual education and training, eliminate barriers and distrust to create an organizational culture that fosters teamwork R Repeatability is the variation in measurements obtained with one instrument used several times by one appraiser on a single parts on characteristic RTY reveals losses from rework
Resolution IV: Two-factor interactions are aliased with three-factor interactions but main effects are free of aliasing. , any two effects for which the number of factors together sum to four or higher may be aliased. For instance, effects AB and CD have two factors each which add up to four, hence these may be aliased.
Residuals are normally within two standard deviations, normally distribution, sum of residuals is zero, equal to the predicted value subtracted from the observed value
RPN conveys the overall risk of a customer problem. Hence, the process element with the highest RPN would be the highest priority for improvement.
Regression Analysis-non linear regressions can explain curvature with more statistical confidence than linear regressions (Simple Linear, Quadratic, Cubic, Multiple Linear, Logarithmic) X's are indpendent, X's are not correlated, no interations,
R-squared expresses the percent of variation explained by the model (0.74 x 0.74 = 0.55
S Sigma is a symbol for Standard Deviation 5S=Seiri, Seiton, seiso, seiketsu, shitsuke Sigma Level concept used to describe a performance of a process relative to a specification limit. Primary goal is to have support and leadership of top management
sequential process flow, the process step that has the longest cycle time is the physical constraint on the entire process. Six Sigma - Control Process Cycle Time; Lean - Reduce Process Variability
Standard Work: takt time
Scatter Plot is used to display the relationship between two variables and is most frequently used in the Analyze phase, although it can also be used at other points in the DMAIC process.
Seasonal is not likely to occur in sequential repetitions of a process over a short period
Standard Deviation is distance between mean and the point of inflection on the normal curve
Supporting Service Requirement: Order-entry is efficient and customer-friendly.
SD = sqrt { sum(x - mu)^2 / (n-1) }
System Diagrams are used to identify the indirect effects and understand the potential for u nintended consequences of dynamic forces acting upon the process.
Subgroups- XBar and R charts Soft Savings- labor and scrap reductions Service and repair policies- most influence on customer of quality Stages of team growth: forming, storming, norming, and performing T T-Test compares a mean against a target, means from two different samples, paired data: the population variance is not known, this is a test of a single population mean with unknown variance,
which can be tested using the t-test for one mean. 10% = 1.318
Team Leader- heavier in context, describe reasoning behind feedback, actionable guidance Three factional factorial:The regions of the customer database are not of interest in the study, but they may
influence the selection of a particular type of troubleshooting procedure and level of support over another due to differences in configuration, energy usage and the like. Blocking can be used to account for the variation due to regions in order to minimize its impact on the total variation.
Total Productive Maintenance: enhances long term availability of a machine Theory of Constraints: most restraints are the results of policies Tree Diagram- reveal the true level of a problems complexity, second step: review diagram for logical flow and completeness
Taguchi's Loss Function: loss to society from poor quality increases as a function of process variability
Transformations are used to reduce the higher order terms in a model
Trend charts are used to display process performance data over time in order to identify any trends in the data.
U U chart defects on different size samples, The u Chart would be correct because it measures defects
per unit (errors per patient) and accommodates a varying sample size. If a constant sample size were used, the X and Moving Range chart would be appropriate. The Upper Control Limit (UCL) of an XmR chart using the median moving range is: X-bar plus 3.14 times the median moving range The Upper Control Limit is: p-bar plus 3 times the square root of ((p-bar times ( 1 minus p-bar))divided by n)
V Variance-how tightly all the various outcome are clustered around the average Variance- measurement and process variance Value added activities: any process (not setup, storage, and inspection) W X
X-bar chart indicates the process improvement action was effective. The eight points below the R-bar line indicate lowered variability in the process which may be a favorable outcome of the improvement action or due to a special cause requiring further investigation.
Y Y=f(x)= Reducing the function of a process input (X) will increase process output (Y). Z