ANOVA - Analysis of Variance – A method for identifying differences in mean values using the variation of the measurements.

Acceptance Region – The region of values for which the null hypothesis is accepted.

Acceptance Sampling – refers to a sampling inspection (as opposed to 100% inspection) in which decisions are made to accept or not to accept a product or service.

Accuracy – is a characteristic of measurement, which addresses how close an observed value is to the true value. It answers the question, “Is it right?”

Affinity Diagram – is a management and planning tool used to organize ideas into natural groupings in a way that stimulates new, creative ideas.

Alpha Risk – The probability of accepting the alternate hypothesis when, in reality, the null hypothesis is true.

Alternate Hypothesis - A tentative explanation, which indicates that an event does not follow a chance distribution; a contrast to null hypothesis.

Arrow Diagram – is a management and planning tool used to develop the best possible schedule and appropriate controls to accomplish the schedule; the critical path method and the program evaluation and review technique.

Assignable Cause – A source of variation which is non random; a change in the source will produce a significant change of some magnitude in the (dependent variable) response.

Attribute Data – are data, which are countable, such as number of rejects or number of errors.

Attribute Sampling Plan – is a plan that allows users to count the number of confirming or non-confirming parts and look for defects. The four types of attribute plans include single, double, multiple, and sequential.

Background Variables – Variables that are of no experimental interest and are not held constant. Their effects are often assumed insignificant or negligible, or they are randomized to ensure that contamination of the primary response does not occur.

Balance Design – A design with an equal number of experimental units in each treatment combinations or run

BB – Black Belt – A Six Sigma project leader

Beta Risk – The probability of accepting the null hypothesis when, in reality, the alternate hypothesis is true.

Black Belt Certification – Recognition obtained upon satisfactory completion of 1 major, or 2 small projects.

Block – Group of homogeneous experimental units.

Blocking – An experiment in which the trials &e made in some restricted order or under restricted conditions.  The experiment is designed such that any nuisance factors do not confuse the true effects f the factors of interest.

Breakthrough – is a method of solving chronic problems, which results from the effective execution of a strategy designed to reach the next level of quality. Such change often requires a paradigm shift within the organization.

C-Charts – Charts, which display the number of defects per sample.

Capability – The ability of a process to stay within specifications and on target.

Capability Ratio (Cp) – is equal to the specification tolerance width divided by the process capability.

Cause & Effect Diagrams – A diagram to show the relationship between and effect (outcome) and its possible causes.  Often displayed with five spines to group the potential causes as method, material, people, environmental and machine.  Asking five “whys” helps to lead toward the root cause.

Center points – Experimental runs with all factor levels set half way between the low and high settings.  Obviously can only be done with quantitative factors.

Central Tendency – Numerical average; e.g. mean, median, and mode; center line on a statistical process control chart.

Champion – is an individual who has an accountability and responsibility for many processes or who is involved in making strategic level decisions for the organization. The champion ensures ongoing dedication of project resources and monitors strategic alignment (also referred to as a sponsor).

Completely Randomized Design (CRD) – An experiment in which one factor of interest is investigated (at multiple levels).   The trials are made in completely randomized order to limit the effect of uncontrollable factors.

Confidence Interval – the range of response values for which one is some percent confident that the true average of the response will fall within that interval that percent of the time.

Confounding – One or more effects that cannot unambiguously be attributed to a single factor or interaction.

Confounded Effects – Effects, which cannot be estimated independently of each other.

Consumer’s Risk – for a sampling plan refers to the probability of acceptance of a lot, the quality of which is designated numerical value representing a level that is seldom desirable. Usually the designated value will be the limiting quality level.

Continuous Probability Distribution – means that the greatest number of observations fall in the center with fewer and fewer observations falling on either side of the average, forming a normal bell-shaped curve.

Control Chart – A methodology for identifying when a process is operating “in Control” (within known statistical boundaries).

Correlation – refers to the measure of the relationship between two sets of numbers or variables.

Correlation Coefficient – describes the magnitude and direction of the relationship between two variables.

Cost Curve – is a model, which shows how developments in new technology, automation and other areas have resulted in the ability to achieve perfection at finite costs.

Cp – Potential Capability Index – Cp = Tolerance /6s.

Cpk – Performance Capability Index – Cpk = min of (USL – mean) or (mean – LSL)/3s.

Cross Functional Team – is a group organized by management and drawn from a variety of functional areas whose responsibility is to identify, analyze and solve chronic problems that are beyond the scope of a quality circle’s effort.

CTQ – Critical to Quality

CTQ 4-Block – Powerful device within the Jugular process for analyzing most critical parameters/process steps vs. current knowledge.

CTQ Flowdown – A very rigorous methodology for allocating requirements and assessing capabilities of the most critical segments of a product prior to M1.

CTQ Jugular – A process of structured and rigorous brainstorming and identification of parameters that the most r\critical together with assessment of current knowledge about the process, culminating in a monitoring devise for maintained success.

Cycle Time – refers to the time that it takes to complete a process from beginning to end and is a critical MBNQA criterion.

Data – Factual information used as a basis for reasoning, discussion, or calculation; often refers to quantitative information.

Degrees of Freedom – Values used in the analysis of variance.  The number of independent pieces of information used to estimate the variability of a factor.

Degrees of Freedom for Error – Values used in the analysis of variance to estimate the process noise.  Without a good estimate of the process noise, determination of which factors are significant and to what degree may be fruitless.  A rule f thumb is 5 degrees of freedom for error at a minimum.  This can equate to at least six replicates.

Defect – Any parameter identified to be evaluated to a given standard, which fails to meet that standard.

Defective (part) –A part identified to be evaluated to a given standard, which fails to meet any portion of that standard.  A single defective part could have multiple defects.

Demographics – are variables among buyers in the consumer market, which include geographic location, age, sex, marital status, family size, social class, education, nationality, occupation and income.

Density Function – The function that yields the probability that a particular random variable takes on any one of its possible values.

DFT – Demand Flow Technology – Materials Management methodology that assures adequate, but not excessive availability of material at the specific time of the need – neither early nor late.

Discrete Probability Distribution – means that the measured process variable takes on a finite or limited number of values; no other possible values exist.

Discrete Random Variable – A random variable that can assume values only from a definite number of discrete values.

Distributions – Tendency of large number of observations to group themselves around some central value with a certain amount of variation on either side.

DOA – Dead on Arrival –Product, which will not work upon customer’s receipt. 

DoE – Design of Experiments –Any of a class of matrices (usually orthogonal) used to understand high-contribution factors. Most often associated with factorial designs.

DPU –Defects Per Unit – #defects found / # Total units physically evaluated.

DPPM – Defective Parts Per Million (outside of specification) – (# defective units /#total units x 1,000,000.

Duncan’s Method – a statistical method used to determine which levels of a factor cause a change in the response.  Used only after the ANOVA indicates a difference among all levels.

Effect – the average change in the response when a factor is changed from a low level to a high level.

Error – the inherent variability in a process.  Represents the change in a response when no change in the factor is made.  See noise.

Estimate – a prediction of some response based on the level of impact of all factors in a process.  See prediction.

EVOP – Evolutionary Operation – A method of conducting designed experiments on an ongoing process without interrupting affecting its efficiency.

EWMA – Exponentially Weighted Moving Average – a control charting methodology that utilizes historical data at an exponentially diminishing weighted value.

Experimental Region – All possible factor-level combinations for which experimentation is possible.  Also known as Factor Space

Experimental Unit – The unit that is observed and measured during the experiment.  Also known as unit of analysis.

External Failure Costs – are costs associated with defects found after the customer receives the product or service.

F Test – A statistical test to determine if a difference exists between two variances.

Factor – an input to a process, which can be changed during experimentation.  Can qualitative (e.g. type of additive) or quantitative (e.g. temperature, pressure)

Factor Analysis – is a statistical technique that examines the relationships between a single dependent variable and multiple independent variations. For example, it is used to determine which questions on a questionnaire are related to specific question such as, “Would you buy this product again?”

Factor, Fixed – If factor levels are specifically assigned, the factor is said to be fixed. Inferences generalize to only those factors.  Effects are of interest.

Factor, Monitored – a factor (usually uncontrollable and hence cannot be held constant) that is observed throughout the experiment and can possibly be correlated to part of the unexplained variation in the process.

Factor, Nuisance – a factor that is known to cause variability in the process; it is not desired to investigate the factor, but rather not to have the factor influence the effect of the factor of interest.  See blocking.

Factor Random – If factor levels are selected randomly from a population of values, the factor is said to be Random.  Variance components are of interest.

Full Factorial Experiment – A class of DoE where 2 levels of several variables are only partially explored.  Used to screen out ‘Trivial many’ and allow focus on ’vital few’ variables controlling the process.

Fixed Effects Factor – a factor for which the levels are chosen selectively.  For example, the effect of temperature will be investigated at 400, 450 and 500 degrees (Compare with Random effects Factor).

Fluctuations – Variances in data, which are caused by a large number of minute variation differences.

Fractional 2k Designs – all factors are run at a low level and high level (see Fractional Factorial Designs).

Fractional 3K Designs – all factors are run at three levels: a low, medium and high (see Fractional Factorial Designs).

Frequency Distribution – The pattern or shape formed by the group of measurements in a distribution.

Gage R&R – Gage Repeatability and Reproducibility.  An analysis of the percent of total variation of a distribution that can be attributed to variation in the measurement system.

Gage Repeatability – Variation in the measurements obtained when one operator uses the same gage for measuring the identical characteristics of the same parts.

Gage Reproducibility – Variation in the average of measurements made by different operators using the same gage when measuring identical characteristics of the same parts.

Gantt Chart – is a project management technique by which the activities of a project are displayed graphically and in sequential order and are plotted against time.

Gap Analysis – is a technique that compares a company’s existing state to its desired state (as expressed by its long term plans) and determines what needs to be done to remove or minimize the gap.

GLM – General Liner Model – A form of ANOVA that allows for a small degree of unbalance in the experimental design.

Goal – is a non-quantitative statement of general intent, aim, or desire; it is the end point toward which management directs its efforts and resources.

Graeco – Latin-Square Design – an experimental design in which one factor of interest is investigated and three nuisance factors are blocked against.

HALT – Highly Accelerated Life Testing – One of several methods for achievement of a reliable design.  Concept is to test a product to extreme (failure) conditions, find root cause of failure, improve design, and repeat the process.

Hetero-skedasticity unequal variances – This condition when applied to factor levels may affect the conclusion from ANOVA.

Histogram – A bar chart to show the distribution of the collected data.

Hoshin Planning – is a methodology for organizing and focusing enterprise efforts on critical issues impacting its success.

Hyper-Graeco-Latin Square Design – an experimental design in which one factor of interest is investigated and four nuisance factors are blocked against.

Hypothesis – an assertion that is tested using a statistical technique.  The hypothesis will either be rejected or insufficient evidence will be available to reject.

Interrelationship Diagram – is a management and planning tool that displays the relationship between factors in a complex situation. It identifies meaningful categories from a mass of ideas and is useful when relationship is difficult to determine.

Interaction – a condition in which the effect of the level of a factor on a response is different for a different levels of a second factor.  There are two-way interactions, three-way interactions, etc.

Instability – Unnaturally large fluctuations in a pattern.

Interval – Numeric categories with equal unit of measure by no absolute zero point, i.e. quality scale index.

IX-MR – Individual X and Moving Range – a control chart of sequential data points, together with a chart of ranges between points.

Just-in-time Training – is training that is offered to employees, as it is needed, so that employees will be able to use their new skills immediately after training.

Kano Model – is a representation of the three levels of customer satisfaction defined as dissatisfaction, neutrality, and delight.

Latin Square Design (LSD) – an experimental design in which one factor of interest is investigated and two nuisance factors are blocked against.

Line Charts – Charts used to track the performance without relationship to the process capability or control limits.

Level – A setting or value of a factor.  Can be qualitative (e.g. additive A and additive B) or qualitative (e.g. 1000psi, 2000 psi.)

Main Effect – The change in response that occurs when a factor is changed from its low level to its high level.

Matrix Chart – is a management and planning tool that shows the relationship among various groups of data; it yields information about the relationships and the importance of task/method elements of the subjects.

MBB – Master Black Belt – A Six Sigma trainer and project mentor.

MBQNA – Malcom Baldrige National Quality Award – is an award that recognizes American companies for business performance excellence and quality achievements. The award criterions describe a total quality management system and include an approach, deployment characteristics, and results that can be applied to the development of quality system.

Mean – Measure of central tendency of a variable – the first moment around the origin.

Mean Square – a column in the ANOVA table that represents the variance of response due to different sources of variability.

Mean Square Error – an entry in the ANOVA table that represents the variance of a response at a given levels of all factors.  An estimate of the variance of a response due to noise (error).

Median – is the middle number or the center value of a set of data when all the data are arranged in an increasing sequence.

Mode – is the score that occurs most frequently in the data.

Minitab – GE’s current statistical analysis software application of choice.

Multiple Comparison Procedure – a statistical method used to determine which levels of a factor cause a change in the response.  Used only after the ANOVA Indicates a difference among all levels.  Examples are Fisher’s method, Duncan’s method, and Scheffe’s method.

Multi-Vari Analysis – A graphical method of decomposing the sources of process variation into their basic components.  This technique is an early step in removing some of the trivial many and preparing a sub-set of factors for designed experimentation.

Multivariate Statistical Methods – Statistical tools for analyzing a set of variables to determine their influence on several responses.  Includes a wide class of statistical tools such as regression, principle components, factor analysis, clustering, and discriminent analysis.

Nominal Group Technique – is a problem solving technique used to generate ideas related to a particular subject. Team members write down their ideas individually and share them one at a time. When all ideas are recorded, they are discussed and prioritized by the group.

Nested Design – an experimental design in which a one factor has different level settings depending on the level of another variable.  For example, batches within different suppliers, levels of competing additives, etc.

Noise – the inherent variability in a process.  Represents the change in a response when no change in the factor is made. See error.

Nominal – Unordered categories which indicate membership or non-membership with no implication of quality, i.e. assembly area number 1, part numbers, etc.

Nonconformity – A condition within a unit, which does not confirm to some specific specification/standard and or requirement.

Normal Distribution – a bell-shaped curve of probabilities that describe many natural processes.  Can occur also in situations in which replicates are taken and are averaged

Normal Probability Plot – a graphical method for investigating whether a sample might have come from a population with a normal distribution.  Often used to check the validity of using ANOVA.

One-Way ANOVA – analysis of variance for investigating a single factor at multiple levels. See ANOVA.

Optimization – Finding the combination of factors and levels that produces the most desirable output from a process.

Pareto Chart – A bar chart to display events with respect to a common metric (#of times, $, time, etc.)

Ordinal – Ordered categories (ranking) with no information about distance between each category.

P Chart – Charts used to plot percent defectives in a sample.

Parameter – A constant defining a particular property of the density function of a variable.

Pareto Chart – A bar chart to display events with respect to a common metric

Plackettt-Burman Design – a designed experiment used in screening experimentation in which a minimal number of trials are need.  Typically only one main effects are investigated with no estimate of the interaction effects.

Plan-do-check-act Cycle – is a continuous improvement model that teaches that organizations should plan an action, do it, check to see how it confirms to plan and expectations, and act on what has been learned.

Point Estimate – the best single value estimate of some prediction or mean response – should be used in conjunction with confidence and/or prediction intervals.

Poka – Yoke – is a term that means to foolproof the process by building safeguards into the system that avoid or immediately find errors.

Population – A group of similar items from which a sample is drawn. Often referred to as the universe.

Pre-control – A methodology for establishing statistically sound probabilities of goodness when a process is starting.

PRD – Phase Review Discipline – A rigorous methodology for new product introduction, which includes milestones at critical points.

Precision – is a characteristic of measurement, which addresses the consistency or repeatability of a measurement system when the identical item is measured a number of times.

Prediction – a best estimate of some response for a given set of levels for all factors.

Prediction Interval – the range of values for a response of which one is some percent confident that a future observation will fall within.  See Confidence Interval.

Prevention Costs – are costs incurred to keep internal and external failure costs and appraisal costs to a minimum.

Prioritization matrix – is a management and planning tool used to determine the highest-priority options/alternatives to accomplish an objective.

Process Capability – refers to the limits within which a tool or process operates, based upon minimum variability as governed by the prevailing circumstances.

Process Decision Program Charts (PDPC) – is a management and planning tool that identifies all events that can go wrong and the appropriate counter measures for these events. It graphically represents all sequence that lead to a desirable effect.

Process Demographics – The list of conditions/states of various factors during the time a response is generated.  These help us to understand what area of the process may be driving the problem.

Process Mapping – is the flowcharting of a work process.

Producer’s Risk – for a sampling plan refers to the probability of not accepting a lot, the quality of which has a designated numerical value representing a level that is generally desirable. Usually the designated value will be the acceptable quality level.

Project Life Cycle – refers to the four sequential phases of project management: conceptualization, planning, implementation, and completion.

Quality – denotes an excellence in goods and services, especially to the degree they confirm to requirements and satisfy customers.

Quality Control – is the operational techniques and activities that are used to fulfill requirements for quality and is aimed at both monitoring a process and eliminating the causes of unsatisfactory process.

Quality Function – is the entire collection of activities through which we achieve fitness-for-use, no matter where these activities are performed.

Quality Function Deployment – is a process used to understand the voice of the customer and to translate customer expectations into technical design parameters for each stage of the product development cycle.

Quality System – is the organizational structure, procedure, processes, and resources needed to implement quality management.

Random Sampling – is a sampling method in which every element in the population has an equal chance of being included.

Random Effects Factor – a factor for which the levels are chosen at random from a definable population.  For example, the effect of batches will be investigated by randomly choosing five batches.  (Compare with Fixed Effects Factor).

Randomization – Mixing up the order of the runs in an experiment as completely as is practical.

Randomized Block Design – an experiment in which one factor of interest is investigated and one nuisance factor is blocked against.

Regression Analysis – is a study used to understand the relationship between two or more variables. Regression analysis makes it possible to predict one variable from the knowledge about the other.

Reliability – refers to the ability of a feedback instrument to produce the same results over the repeated administration.

Repetition – Running several experimental unties over one treatment combination. Contrast with Replication.

Replication – repeated runs at the same experimental conditions; provides and estimate of the noise in the process.

Residuals – the difference between the observed response and the predicted response for a given set of factor conditions.  Used in model validation and process investigation.

Resolution – a description of fractional factorial designs that gives the degree to which factors will be aliased with other factors’ interactions.

Response – an output from the process, which will be measured during the experiment.

Root Cause Analysis – is a quality tool used to distinguish the source of defects or problems. It is a structured approach that focuses on the decisive or original cause of a problem or condition.

RSM – Response Surface Methodology – a class of designed experiments where curvature of the vital few is examined and understood.  Subsets include central composite designs with star or face points.

R-Square – percent of variability in the response explained by the controlled factors. 

Run – a set of process conditions defined by specifying levels of all the factors in the experiment. Also knows as a Treatment Combination.

Run chart – a sequential time series plot of data, which provides some statistical analysis capabilities and probabilities.

Sample – One or more observations drawn from a larger collection of observations or universe.

Scatter Plot – a chart (dot plot) to show the relationship between two variables.

SCBN – Supplier Change Notice – The communication device for requesting a change to a purchased part, initiated either by a Supplier to GEMS, or by GEMS to a Supplier.

Screening Experiment – a technique used to characterize a process (usually assumes liner changes in the response for a change of factor levels) (Compare with RSM).

Signal to Noise Ration – a ratio that depends on the variability in the response due to changing factor levels relative to the variability when there is no change in the factor level.

(Six) 6-s process – a stable process operating such that it’s output has minimum Cpk of 1.5.

Skewness – a condition in which the normal distribution is shifted to the left or right, that is, no longer symmetric.  Can influence the validity of using ANOVA.

SPC – Statistical Process Control – Used to monitor process stability preferably after modification to desired state.

Special Cause of Variation – are the factors that disrupt the usual flow of work. Processes with special causes are unstable and unpredictable.

Standard Deviation – A statistical index of variability, which describes the spread of the data.

Statistical Control – A quantitative condition, which describes a process free of assignable/special causes of variation.

Statistical Process Control – The application of statistical methods and procedures relative to a process and given set of standards.

SWOT Analysis – is an assessment of an organization’s key strengths, weaknesses, opportunities, and threats. It considers factors such as the organization’s industry, the competitive position, functional areas and management.

T-Test – A statistical comparison of man values of a sample, assuming a normal population.

Test of Significance – A procedure to determine whether a quantity subjected to random variation differs from a postulated value by an amount greater than that due to random variation alone.

Tree Diagram – is a management and planning tool that shows the complete range of subtasks required to achieve an objective. A problem solving method can be identified from this analysis.

Trivial Many – The factors that have long been thought to have some influence on the process but really account for very little of the variation in performance.

Total Quality Management – is a strategic, integrated management system for achieving customer satisfaction, which involves all managers and employees and uses quantitative methods to continuously improve an organizations process.

Treatment Combination – a set of process conditions defined by specifying levels of all the factors in the experiment.  Also known as a Run.

Two-Way ANOVA – analysis of variance for investigating two factors at multiple levels

Two –way Interaction Plot – a scatter plot of the average responses (vertical axis) as a factor of one factor (horizontal axis) and the average response for each level of the second factor connected by lines.

Type I Error – the error of assuming that the hypothesis s false when in fact it is true.  Associated probability is labeled a.

Type II Error – the error of assuming that the hypothesis is true when it is in fact false.  Associated probability is labeled b.

Unbalanced Design – A design with an unequal number of experimental units in each treatment combinations or run.

USL or LSL – Upper (Lower) Specification Limit – boundaries of design criteria or statistical boundaries of a control chart.

UCL or LCL - Upper (Lower) Control Limit – the statistical boundaries of a control chart OR Upper (Lower) Confidence Limit – (used in T-test, ANOVA, etc.).

Variable – A characteristic that may take on different values.

Variance – Provides a measure of dispersion.  The square root of the variance is the standard deviation.  The 2^nd moment around the mean.

Variation – Any quantifiable difference between individual measurements; such differences can be classified as being due to common (random) or special (assignable) causes.

Vital Few – The factors that are critical in controlling the process.

Z_B– Z benchmark, Z_st, short-term sigma – Assumes process is centered (on target) with short term variation.

Z value – Calculation of how many sigmas fit between the process output average and the closest specification limit.