This Glossary reflects those Estimating Terms that are either in common usage or have been defined for the purposes of this series of guides. Not all the terms are used in every volume, but where they do occur, their meaning is intended to be consistent.
3-Point Estimate A 3-Point Estimate is an expression of uncertainty around an Estimate Value. It usually expresses Optimistic, Most Likely and Pessimistic Values.
Accuracy Accuracy is an expression of how close a measurement, statistic or estimate is to the true value, or to a defined standard.
Actual Cost (AC) See Earned Value Management Abbreviations and Terminology
ACWP (Actual Cost of Work Performed) or Actual Cost (AC) See Earned Value Management Terminology
Additive/Subtractive Time Series Model See Time Series Analysis
Adjusted R-Square Adjusted R-Square is a measure of the “Goodness of Fit” of a Multi-Linear Regression model to a set of data points, which reduces the Coefficient of Determination by a proportion of the Unexplained Variance relative to the Degrees of Freedom in the model, divided by the Degrees of Freedom in the Sum of Squares Error.
ADORE (Assumptions, Dependencies, Opportunities, Risks, Exclusions) See Individual Terms.
Alternative Hypothesis An Alternative Hypothesis is that supposition that the difference between an observed value and another observed or assumed value or effect, cannot be legitimately attributable to random sampling or experimental error. It is usually denoted as H1.
Analogous Estimating Method or Analogy See Analogical Estimating Method.
Analogical Estimating Method The method of estimating by Analogy is a means of creating an estimate by comparing the similarities and/or differences between two things, one of which is used as the reference point against which rational adjustments for differences between the two things are made in order establish an estimate for the other.
Approach See Estimating Approach.
Arithmetic Mean or Average The Arithmetic Mean or Average of a set of numerical data values is a statistic calculated by summating the values of the individual terms and dividing by the number of terms in the set.
Assumption An Assumption is something that we take to be broadly true or expect to come to fruition in the context of the Estimate.
Asymptote An Asymptote to a given curve is a straight line that tends continually closer in value to that of the curve as they tend towards infinity (positive or negative). The difference between the asymptote and its curve reduces towards but never reaches zero at any finite value.
AT (Actual Time) See Earned Value Management Abbreviations and Terminology.
Average See Arithmetic Mean.
Average (Mean) Absolute Deviation (AAD) The Mean or Average Absolute Deviation of a range of data is the average ‘absolute’ distance of each data point from the Arithmetic Mean of all the data points, ignoring the sign depicting whether each point is less than or greater than the Arithmetic Mean.
Axiom An Axiom is a statement or proposition that requires no proof, being generally accepted as being self-evidently true at all times.
BAC (Budget At Completion) See Earned Value Management Abbreviations and Terminology.
Base Year Values ‘Base Year Values’ are values that have been adjusted to be expressed relative to a fixed year as a point of reference e.g., for contractual price agreement.
Basis of Estimate (BoE) A Basis of Estimate is a series of statements that define the assumptions, dependencies and exclusions that bound the scope and validity of an estimate. A good BoE also defines the approach, method and potentially techniques used, as well as the source and value of key input variables, and as such supports Estimate TRACEability.
BCWP (Budgeted Cost of Work Performed) See Earned Value Management Abbreviations and Terminology .
BCWS (Budgeted Cost of Work Scheduled) See Earned Value Management Abbreviations and Terminology.
Benford’s Law Benford’s Law is an empirical observation that in many situations the first or leading digit in a set of apparently random measurements follows a repeating pattern that can be predicted as the Logarithm of one plus the reciprocal of the leading digit. It is used predominately in the detection of fraud.
Bessel’s Correction Factor In general, the variance (and standard deviation) of a data sample will understate the variance (and standard deviation) of the underlying data population. Bessel’s Correction Factor allows for an adjustment to be made so that the sample variance can be used as an unbiased estimator of the population variance. The adjustment requires that the Sum of Squares of the Deviations from the Sample Mean be divided one less than the number of observations or data points i.e. n-1 rather than the more intuitive the number of observations. Microsoft Excel takes this adjustment into account.
Bottom-up Approach In a Bottom-up Approach to estimating, the estimator identifies the lowest level at which it is appropriate to create a range of estimates based on the task definition available, or that can be inferred. The overall estimate, or higher level summaries, typically through a Work Breakdown Structure, can be produced through incremental aggregation of the lower level estimates. A Bottom-up Approach requires a good definition of the task to be estimated, and is frequently referred to as detailed estimating or as engineering build-up.
Chauvenet’s Criterion A test for a single Outlier based on the deviation Z-Score of the suspect data point.
Chi-Squared Test or χ2-Test The Chi-Squared Test is a “goodness of fit” test that compares the variance of a sample against the variance of a theoretical or assumed distribution.
Classical Decomposition Method (Time Series) Classical Decomposition Method is a means of analysing data for which there is a seasonal and/or cyclical pattern of variation. Typically, the underlying trend is identified, from which the average deviation or variation by season can be determined. The method can be used for multiplicative and additive/ subtractive Time Series Models.
Closed Interval A Closed Continuous Interval is one which includes its endpoints, and is usually depicted with square brackets: [Minimum, Maximum].
Coefficient of Determination The Coefficient of Determination is a statistical index which measures how much of the total variance in one variable can be explained by the variance in the other variable. It provides a measure of how well the relationship between two variables can be represented by a straight line.
Coefficient of Variation (CV) The Coefficient of Variation of a set of sample data values is a dimensionless statistic which expresses the ratio of the sample’s Standard Deviation to its Arithmetic Mean. In the rare cases where the set of data is the entire population, then the Coefficient of Variation is expressed as the ratio of the population’s Standard Deviation to its Arithmetic Mean. It can be expressed as either a decimal or percentage.
Collaborative Working Collaborative Working is a term that refers to the management strategy of dividing a task between multiple partners working towards a common goal where there a project may be unviable for a single organisation. There is usually a cost penalty of such collaboration as it tends to create duplication in management and in integration activities.
Collinearity & Multicollinearity Collinearity is an expression of the degree to which two supposedly independent predicator variables are correlated in the context of the observed values being used to model their relationship with the dependent variable that we wish to estimate. Multicollinearity is an expression to which collinearity can be observed across several predicator variables.
Complementary Cumulative Distribution Function (CCDF) The Complementary Cumulative Distribution Function is the theoretical or observed probability of that variable being greater than a given value. It is calculated as the difference between 1 (or 100%) and the Cumulative Distribution Function, 1-CDF.
Composite Index A Composite Index is one that has been created as the weighted average of a number of other distinct Indices for different commodities.
Concave Curve A curve in which the direction of curvature appears to bend towards a viewpoint on the x-axis, similar to one that would be observed when viewing the inside of a circle or sphere.
Cone of Uncertainty A generic term that refers to the empirical observation that the range of estimate uncertainty or accuracy improves through the life of a project. It is typified by its cone or funnel shape appearance.
Confidence Interval A Confidence Interval is an expression of the percentage probability that data will lie between two distinct Confidence Levels, known as the Lower and Upper Confidence Limits, based on a known or assumed distribution of data from either a sample or an entire population.
See also Prediction Interval.
Confidence Level A Confidence Level is an expression of the percentage probability that data selected at random from a known or assumed distribution of data (either a sample or an entire population), will be less than or equal to a particular value.
Confidence Limits The Lower and Upper Confidence Limits are the respective Confidence Levels that bound a Confidence Interval, and are expressions of the two percentage probabilities that data will be less or equal to the values specified based on the known or assumed distribution of data in question from either a sample or an entire population.
See also Confidence Interval.
Constant Year Values ‘Constant Year Values’ are values that have been adjusted to take account of historical or future inflationary effects or other changes, and are expressed in relation to the Current Year Values for any defined year. They are often referred to as ‘Real Year Values’.
Continuous Probability Distribution A mathematical expression of the relative theoretical probability of a random variable which can take on any value from a real number range. The range may be bounded or unbounded in either direction.
Convex Curve A curve in which the direction of curvature appears to bend away from a viewpoint on the x-axis, similar to one that would be observed when viewing the outside of a circle or sphere.
Copula A Copula is a Multivariate Probability Distribution based exclusively on a number Uniform Marginal Probability Distributions (one for each variable).
Correlation Correlation is a statistical relationship in which the values of two or more variables exhibit a tendency to change in relationship with one other. These variables are said to be positively (or directly) correlated if the values tend to move in the same direction, and negatively (or inversely) correlated if they tend to move in opposite directions.
Cost Driver See Estimate Drivers.
Covariance The Covariance between a set of paired values is a measure of the extent to which the paired data values are scattered around the paired Arithmetic Means. It is the average of the product of each paired variable from its Arithmetic Mean.
CPI (Cost Performance Index) See Earned Value Management Abbreviations and Terminology.
Crawford’s Unit Learning Curve A Crawford Unit Learning Curve is an empirical relationship that expresses the reduction in time or cost of each unit produced as a power function of the cumulative number units produced.
Critical Path The Critical Path at a point in time depicts the string of dependent activities or tasks in a schedule for which there is no float or queuing time. As such the length of the Critical Path represents the quickest time that the schedule can be currently completed based on the current assumed activity durations.
Cross-Impact Analysis A Cross-Impact Analysis is a qualitative technique used to identify the most significant variables in a system by considering the impact of each variable on the other variables.
Cumulative Average A Point Cumulative Average is a single term value calculated as the average of the current and all previous consecutive recorded input values that have occurred in a natural sequence.
A Moving Cumulative Average, sometimes referred to as a Cumulative Moving Average, is an array (a series or range of ordered values) of successive Point Cumulative Average terms calculated from all previous consecutive recorded input values that have occurred in a natural sequence.
Cumulative Distribution Function (CDF) The Cumulative Distribution Function of a Discrete Random Variable expresses the theoretical or observed probability of that variable being less than or equal to any given value. It equates to the sum of the probabilities of achieving that value and each successive lower value.
The Cumulative Distribution Function of a Continuous Random Variable expresses the theoretical or observed probability of that variable being less than or equal to any given value. It equates to the area under the Probability Density Function curve to the left of the value in question.
See also the Complementary Cumulative Distribution Function.
Current Year (or Nominal Year) Values ‘Current Year Values’ are historical values expressed in terms of those that were current at the historical time at which they were incurred. In some cases, these may be referred to as ‘Nominal Year Values’.
CV (Cost Variance) See Earned Value Management Abbreviations and Terminology
Data Type Primary Data is that which has been taken directly from its source, either directly or indirectly, without any adjustment to its values or context.
Secondary Data is that which has been taken from a known source, but has been subjected to some form of adjustment to its values or context, the general nature of which is known and has been considered to be appropriate.
Tertiary Data is data of unknown provenance. The specific source of data and its context is unknown, and it is likely that one or more adjustments of an unknown nature have been made, in order to make it suitable for public distribution.
Data Normalisation Data Normalisation is the act of making adjustments to, or categori-sations of, data to achieve a state where data the can be used for comparative purposes in estimating.
Decile A Decile is one of ten subsets from a set of ordered values which nominally contain a tenth of the total number of values in each subset. The term can also be used to express the values that divide the ordered values into the ten ordered subsets.
Degrees of Freedom Degrees of Freedom are the number of different factors in a system or calculation of a statistic that can vary independently.
DeJong Unit Learning Curve A DeJong Unit Learning Curve is a variation of the Crawford Unit Learning Curve that allows for an incompressible or ‘unlearnable’ element of the task, expressed as a fixed cost or time.
Delphi Technique The Delphi Technique is a qualitative technique that promotes consensus or convergence of opinions to be achieved between diverse subject matter experts in the absence of a clear definition of a task or a lack of tangible evidence.
Dependency A Dependency is something to which an estimate is tied, usually an uncertain event outside of our control or influence, which if it were not to occur, would potentially render the estimated value invalid. If it is an internal dependency, the estimate and schedule should reflect this relationship
Descriptive Statistic A Descriptive Statistic is one which reports an indisputable and repeatable fact, based on the population or sample in question, and the nature of which is described in the name of the Statistic.
Discount Rate The Discount Rate is the percentage reduction used to calculate the present-day values of future cash flows. The discount rate often either reflects the comparable market return on investment of opportunities with similar levels of risk, or reflects an organisation’s Weighted Average Cost of Capital (WACC), which is based on the weighted average of interest rates paid on debt (loans) and shareholders’ return on equity investment.
Discounted Cash Flow (DCF) Discounted Cash Flow (DCF) is a technique for converting estimated or actual expenditures and revenues to economically comparable values at a common point in time by discounting future cash flows by an agreed percentage discount rate per time period, based on the cost to the organisation of borrowing money, or the average return on comparable investments.
Discrete Probability Distribution A mathematical expression of the theoretical or empirical probability of a random variable which can only take on predefined values from a finite range.
Dixon’s Q-Test A test for a single Outlier based on the distance between the suspect data point and its nearest neighbour in comparison with the overall range of the data.
Driver See Estimate Drivers.
Earned Value (EV) See Earned Value Management Terminology.
Earned Value Management (EVM) Earned Value Management is a collective term for the management and control of project scope, schedule and cost.
Earned Value Analysis Earned Value Analysis is a collective term used to refer to the analysis of data gathered and used in an Earned Value Management environment.
Earned Value Management Abbreviations and Terminology (Selected terms only) ACWP (Actual Cost of Work Performed) sometimes referred to as Actual Cost (AC) Each point represents the cumulative actual cost of the work completed or in progress at that point in time. The curve represents the profile by which the actual cost has been expended for the value achieved over time.
AT (Actual Time) AT measures the time from start to time now.
BAC (Budget At Completion) The BAC refers to the agreed target value for the current scope of work, against which overall performance will be assessed.
BCWP (Budget Cost of Work Performed) sometimes referred to as Earned Value (EV) Each point represents the cumulative budgeted cost of the work completed or in progress to that point in time. The curve represents the profile by which the budgeted cost has been expended over time. The BCWP is expressed in relation to the BAC (Budget At Completion).
BCWS sometimes referred to as Planned Value (PV) Each point represents the cumulative budgeted cost of the work planned to be completed or to be in progress to that point in time. The curve represents the profile by which the budgeted cost was planned to be expended over time. The BCWS is expressed in relation to the BAC (Budget At Completion).
CPI (Cost Performance Index) The CPI is an expression of the relative performance from a cost perspective and is the ratio of Earned Value to Actual Cost (EV/ AC) or (BCWP/ACWP).
CV (Cost Variance) CV is a measure of the cumulative Cost Variance as the difference between the Earned Value and the Actual Cost (EV – AC) or (BCWP – ACWP).
ES (Earned Schedule) ES measures the planned time allowed to reach the point that we have currently achieved.
EAC (Estimate At Completion) sometimes referred to as FAC (Forecast At Completion) The EAC or FAC is the sum of the actual cost to date for the work achieved, plus an estimate of the cost to complete any outstanding or incomplete activity or task in the defined scope of work
ETC (Estimate To Completion) The ETC is an estimate of the cost that is likely to be expended on the remaining tasks to complete the current scope of agreed work. It is the difference between the Estimate At Completion and the current Actual Cost (EAC – ACWP or AC).
SPI (Schedule Performance Index) The SPI is an expression of the relative schedule performance expressed from a cost perspective and is the ratio of Earned Value to Planned Value (EV/PV) or (BCWP/BCWS). It is now considered to be an inferior measure of true schedule variance in comparison with SPI(t).
SPI(t) The SPI(t) is an expression of the relative schedule performance and is the ratio of Earned Schedule to Actual Time (ES/AT).
SV (Schedule Variance) SV is a measure of the cumulative Schedule Variance measured from a Cost Variance perspective, and is the difference between the Earned Value and the Planned Value (EV – PV) or (BCWP – BCWS). It is now considered to be an inferior measure of true schedule variance in comparison with SV(t).
SV(t) SV(t) is a measure of the cumulative Schedule Variance and is the difference between the Earned Schedule and the Actual Time (ES – AT).
Equivalent Unit Learning Equivalent Unit Learning is a technique that can be applied to complex programmes of recurring activities to take account of Work-in-Progress and can be used to give an early warning indicator of potential learning curve breakpoints. It can be used to supplement traditional completed Unit Learning Curve monitoring.
ES (Earned Schedule) See Earned Value Management Abbreviations and Terminology
Estimate An Estimate for ‘something’ is a numerical expression of the approximate value that might reasonably be expected to occur based on a given context, which is described and is bounded by a number of parameters and assumptions, all of which are pertinent to and necessarily accompany the numerical value provided.
Estimate At Completion (EAC) and Estimate To Completion (ETC) See Earned Value Management Abbreviations and Terminology.
Estimate Drivers A Primary Driver is a technical, physical, programmatic or transactional characteristic that either causes a major change in the value being estimated or in a major constituent element of it, or whose value itself changes correspondingly with the value being estimated, and therefore, can be used as an indicator of a change in that value.
A Secondary Driver is a technical, physical, programmatic or transactional characteristic that either causes a minor change in the value being estimated or in a constituent element of it, or whose value itself changes correspondingly with the value being estimated and can be used as an indicator of a subtle change in that value.
Cost Drivers are specific Estimate Drivers that relate to an indication of Cost behaviour.
Estimate Maturity Assessment (EMA) An Estimate Maturity Assessment provides a ‘health warning’ on the maturity of an estimate based on its Basis of Estimate, and takes account of the level of task definition available and historical evidence used.
Estimating Approach An Estimating Approach describes the direction by which the lowest level of detail to be estimated is determined.
See also Bottom-up Approach, Top-down Approach and Ethereal Approach.
Estimating Method An Estimating Method is a systematic means of creating an estimate, or an element of an estimate. An Estimating Methodology is a set or system of Estimating Methods.
See also Analogous Method, Parametric Method and Trusted Source Method.
Estimating Metric An Estimating Metric is a value or statistic that expresses a numerical relationship between a value for which an estimate is required, and a Primary or Secondary Driver (or parameter) of that value, or in relation to some fixed reference point. See also Factor, Rate and Ratio.
Estimating Procedure An Estimating Procedure is a series of steps conducted in a certain manner and sequence to optimise the output of an Estimating Approach, Method and/ or Technique.
Estimating Process An Estimating Process is a series of mandatory or possibly optional actions or steps taken within an organisation, usually in a defined sequence or order, in order to plan, generate and approve an estimate for a specific business purpose.
Estimating Technique An Estimating Technique is a series of actions or steps conducted in an efficient manner to achieve a specific purpose as part of a wider Estimating Method. Techniques can be qualitative as well as quantitative.
Ethereal Approach An Ethereal Approach to Estimating is one in which values are accepted into the estimating process, the provenance of which is unknown and at best may be assumed. These are values often created by an external source for low value elements of work, or by other organisations with acknowledged expertise. Other values may be generated by Subject Matter Experts internal to the organisation where there is insufficient definition or data to produce an estimate by a more analytical approach.
The Ethereal Approach should be considered the approach of last resort where low maturity is considered acceptable. The approach should be reserved for low value elements or work, and situations where a robust estimate is not considered critical.
Excess Kurtosis The Excess Kurtosis is an expression of the relative degree of Peakedness or flatness of a set of data values, relative to a Normal Distribution. Flatter distributions with a negative Excess Kurtosis are referred to as Platykurtic; Peakier distributions with a positive Excess Kurtosis are termed Leptokurtic; whereas those similar to a Normal Distribution are said to be Mesokurtic. The measure is based on the fourth power of the deviation around the Arithmetic Mean.
Exclusion An Exclusion is condition or set of circumstances that have been designated to be out of scope of the current estimating activities and their output.
Exponential Function An Exponential Function of two variables is one in which the Logarithm of the dependent variable on the vertical axis produces a monotonic increasing or decreasing Straight Line when plotted against the independent variable on the horizontal axis.
Exponential Smoothing Exponential Smoothing is a ‘single-point’ predictive technique which generates a forecast for any period based on the forecast made for the prior period, adjusted for the error in that prior period’s forecast.
Extrapolation The act of estimating a value extrinsic to or outside the range of the data being used to determine that value. See also Interpolation.
Factored or Expected Value Technique A technique that expresses an estimate based on the weighted sum of all possible values multiplied by the probability of arising.
Factors, Rates and Ratios See individual terms: Factor Metric, Rate Metric and Ratio Metric
Factor Metric A Factor is an Estimating Metric used to express one variable’s value as a percentage of another variable’s value.
F-Test The F-Test is a “goodness of fit” test that returns the cumulative probability of getting an F-Statistic less than or equal to the ratio inferred by the variances in two samples.
Generalised Exponential Function A variation to the standard Exponential Function which allows for a constant value to exist in the dependent or predicted variable’s value. It effectively creates a vertical shift in comparison with a standard Exponential Function.
Generalised Extreme Studentised Deviate A test for multiple Outliers based on the deviation Z-Score of the suspect data point.
Generalised Logarithmic Function A variation to the standard Logarithmic Function which allows for a constant value to exist in the independent or predictor variable’s value. It effectively creates a horizontal shift in comparison with a standard Logarithmic Function.
Generalised Power Function A variation to the standard Power Function which allows for a constant value to exist in either or both the independent and dependent variables’ value. It effectively creates a horizontal and/or vertical shift in comparison with a standard Power Function.
Geometric Mean The Geometric Mean of a set of n numerical data values is a statistic calculated by taking the nth root of the product of the n terms in the set.
Good Practice Spreadsheet Modelling (GPSM) Good Practice Spreadsheet Modelling Principles relate to those recommended practices that should be considered when developing a Spreadsheet in order to help maintain its integrity and reduce the risk of current and future errors.
Grubbs’ Test A test for a single Outlier based on the deviation Z-Score of the suspect data point.
Harmonic Mean The Harmonic Mean of a set of n numerical data values is a statistic calculated by taking the reciprocal of the Arithmetic Mean of the reciprocals of the n terms in the set.
Heteroscedasticity Data is said to exhibit Heteroscedasticity if data variances are not equal for all data values.
Homoscedasticity Data is said to exhibit Homoscedasticity if data variances are equal for all data values.
Iglewicz and Hoaglin’s M-Score (Modified Z-Score) A test for a single Outlier based on the Median Absolute Deviation of the suspect data point.
Index An index is an empirical average factor used to increase or decrease a known reference value to take account of cumulative changes in the environment, or observed circumstances, over a period of time. Indices are often used as to normalise data.
Inferential Statistic An Inferential Statistic is one which infers something, often about the wider data population, based on one or more Descriptive Statistics for a sample, and as such, it is open to interpretation . . . and disagreement.
Inherent Risk in Spreadsheets (IRiS) IRiS is a qualitative assessment tool that can be used to assess the inherent risk in spreadsheets by not following Good Practice Spreadsheets Principles.
Interdecile Range The Interdecile Range comprises the middle eight Decile ranges and represents the 80% Confidence Interval between the 10% and 90% Confidence Levels for the data.
Internal Rate of Return The Internal Rate of Return (IRR) of an investment is that Discount Rate which returns a Net Present Value (NPV) of zero, i.e. the investment breaks even over its life with no over or under recovery.
Interpolation The act of estimating an intermediary or intrinsic value within the range of the data being used to determine that value. See also Extrapolation.
Interquantile Range An Interquantile Range is a generic term for the group of Quantiles that form a symmetrical Confidence Interval around the Median by excluding the first and last Quantile ranges.
Interquartile Range The Interquartile Range comprises the middle two Quartile ranges and represents the 50% Confidence Interval between the 25% and 75% Confidence Levels for the data.
Interquintile Range The Interquintile Range comprises the middle three Quintile ranges and represents the 60% Confidence Interval between the 20% and 80% Confidence Levels for the data.
Jarque-Bera Test The Jarque-Bera Test is a statistical test for whether data can be assumed to follow a Normal Distribution. It exploits the properties of a Normal Distribution’s Skewness and Excess Kurtosis being zero.
Kendall’s Tau Rank Correlation Coefficient Kendall’s Tau Rank Correlation Coefficient for two variables is a statistic that measures the difference between the number of Concordant and Discordant data pairs as a proportion of the total number of possible unique pairings, where two pairs are said to be concordant if the ranks of the two variables move in the same direction, or are said to be discordant if the ranks of the two variables move in opposite directions.
Laspeyres Index Laspeyres Indices are time-based indices which compare the prices of commodities at a point in time with the equivalent prices for the Index Base Period, based on the original quantities consumed at the Index Base Year.
Learning Curve A Learning Curve is a mathematical representation of the degree at which the cost, time or effort to perform one or more activities reduces through the acquisition and application of knowledge and experience through repetition and practice.
Learning Curve Breakpoint A Learning Curve Breakpoint is the position in the build or repetition sequence at which the empirical or theoretical rate of learning changes.
Learning Curve Cost Driver A Learning Curve Cost Driver is an independent variable which affects or indicates the rate or amount of learning observed.
Learning Curve Segmentation Learning Curve Segmentation refers to a technique which models the impact of discrete Learning Curve Cost Drivers as a product of multiple unit-based learning curves.
Learning Curve Step-point A Learning Curve Step-point is the position in the build or repetition sequence at which there is a step function increase or decrease in the level of values evident on the empirical or theoretical Learning Curve.
Learning Exponent A Learning Exponent is the power function exponent of a Learning Curve reduction and is calculated as the Logarithmic value of the Learning Rate using a Logarithmic Base equivalent to the Learning Rate Multiplier.
Learning Rate and Learning Rate Multiplier The Learning Rate expresses the complement of the percentage reduction over a given Learning Rate Multiplier (usually 2). For example, an 80% Learning Rate with a Learning Multiplier of 2 implies a 20% reduction every time the quantity doubles.
Least Squares Regression Least Squares Regression is a Regression procedure which identifies the ‘Best Fit’ of a pre-defined functional form by minimising the Sum of the Squares of the vertical difference between each data observation and the assumed functional form through the Arithmetic Mean of the data.
Leptokurtotic or Leptokurtic An expression that the degree of Excess Kurtosis in a probability distribution is peakier than a Normal Distribution.
Linear Function A Linear Function of two variables is one which can be represented as a monotonic increasing or decreasing Straight Line without any need for Mathematical Transformation.
Logarithm The Logarithm of any positive value for a given positive Base Number not equal to one is that power to which the Base Number must be raised to get the value in question.
Logarithmic Function A Logarithmic Function of two variables is one in which the dependent variable on the vertical axis produces a monotonic increasing or decreasing Straight Line, when plotted against the Logarithm of the independent variable on the horizontal axis.
Mann-Whitney U-Test sometimes known as Mann-Whitney-Wilcoxon U-Test A U-Test is used to test whether two samples could be drawn from the same population by comparing the distribution of the joint ranks across the two samples.
Marching Army Technique sometimes referred to as Standing Army Technique The Marching Army Technique refers to a technique that assumes that costs vary directly in proportion with a schedule.
Mathematical Transformation A Mathematical Transformation is a numerical process in which the form, nature or appearance of a numerical expression is converted into an equivalent but non-identical numerical expression with a different form, nature or appearance.
Maximum The Maximum is the largest observed value in a sample of data, or the largest potential value in a known or assumed statistical distribution. In some circumstances, the term may be used to imply a pessimistic value at the upper end of potential values rather than an absolute value.
Mean Absolute Deviation See Average Absolute Deviation (AAD).
Measures of Central Tendency Measures of Central Tendency is a collective term that refers to those descriptive statistics that measure key attributes of a data sample (Means, Modes and Median).
Measures of Dispersion and Shape Measures of Dispersion and Shape is a collective term that refers to those descriptive statistics that measure the degree and/or pattern of scatter in the data in relation to the Measures of Central Tendency.
Median The Median of a set of data is that value which occurs in the middle of the sequence when its values have been arranged in ascending or descending order. There are an equal number of data points less than and greater than the Median.
Median Absolute Deviation (MAD) The Median Absolute Deviation of a range of data is the Median of the ‘absolute’ distance of each data point from the Median of those data points, ignoring the “sign” depicting whether each point is less than or greater than the Median.
Memoryless Probability Distribution In relation to Queueing Theory, a Memoryless Probability Distribution is one in which the probability of waiting a set period of time is independent of how long we have been waiting already. The probability of waiting longer than the sum of two values is the product of the probabilities of waiting longer than each value in turn. An Exponential Distribution is the only Continuous Probability Distribution that exhibits this property, and a Geometric Distribution is the only discrete form.
Mesokurtotic or Mesokurtic An expression that the degree of Excess Kurtosis in a probability distribution is comparable with a Normal Distribution.
Method See Estimating Method.
Metric A Metric is a statistic that measures an output of a process or a relationship between a variable and another variable or some reference point.
See also Estimating Metric.
Minimum The Minimum is the smallest observed value in a sample of data, or the smallest potential value in a known or assumed statistical distribution. In some circumstances, the term may be used to imply an optimistic value at the lower end of potential values rather than an absolute value.
Mode The Mode of a set of data is that value which has occurred most frequently, or that which has the greatest probability of occurring.
Model Validation and Verification See individual terms: Validation and Verification.
Monotonic Function A Monotonic Function of two paired variables is one that when values are arranged in ascending numerical order of one variable, the value of the other variable either perpetually increases or perpetually decreases.
Monte Carlo Simulation Monte Carlo Simulation is a technique that models the range and relative probabilities of occurrence, of the potential outcomes of a number of input variables whose values are uncertain but can be defined as probability distributions.
Moving Average A Moving Average is a series or sequence of successive averages calculated from a fixed number of consecutive input values that have occurred in a natural sequence. The fixed number of consecutive input terms used to calculate each average term is referred to as the Moving Average Interval or Base.
Moving Geometric Mean A Moving Geometric Mean is a series or sequence of successive geometric means calculated from a fixed number of consecutive input values that have occurred in a natural sequence. The fixed number of consecutive input terms used to calculate each geometric mean term is referred to as the Moving Geometric Mean Interval or Base.
Moving Harmonic Mean A Moving Harmonic Mean is a series or sequence of successive harmonic means calculated from a fixed number of consecutive input values that have occurred in a natural sequence. The fixed number of consecutive input terms used to calculate each harmonic mean term is referred to as the Moving Harmonic Mean Interval or Base.
Moving Maximum A Moving Maximum is a series or sequence of successive maxima calculated from a fixed number of consecutive input values that have occurred in a natural sequence. The fixed number of consecutive input terms used to calculate each maximum term is referred to as the Moving Maximum Interval or Base.
Moving Median A Moving Median is a series or sequence of successive medians calculated from a fixed number of consecutive input values that have occurred in a natural sequence. The fixed number of consecutive input terms used to calculate each median term is referred to as the Moving Median Interval or Base.
Moving Minimum A Moving Minimum is a series or sequence of successive minima calculated from a fixed number of consecutive input values that have occurred in a natural sequence. The fixed number of consecutive input terms used to calculate each minimum term is referred to as the Moving Minimum Interval or Base.
Moving Standard Deviation A Moving Standard Deviation is a series or sequence of successive standard deviations calculated from a fixed number of consecutive input values that have occurred in a natural sequence. The fixed number of consecutive input terms used to calculate each standard deviation term is referred to as the Moving Standard Deviation Interval or Base.
Multicollinearity See Collinearity.
Multiplicative Time Series Model See Time Series Analysis.
Multi-Variant Unit Learning Multi-Variant Unit Learning is a technique that considers shared and unique learning across multiple variants of the same or similar recurring products.
Net Present Value The Net Present Value (NPV) of an investment is the sum of all positive and negative cash flows through time, each of which have been discounted based on the time value of money relative to a Base Year (usually the present year).
Nominal Year Values ‘Nominal Year Values’ are historical values expressed in terms of those that were current at the historical time at which they were incurred. In some cases, these may be referred to as ‘Current Year Values’.
Norden-Rayleigh Curve A Norden-Rayleigh is an empirical relationship that models the distribution of resource required in the non-recurring concept demonstration or design and development phases.
Null Hypothesis A Null Hypothesis is that supposition that the difference between an observed value or effect and another observed or assumed value or effect, can be legitimately attributable to random sampling or experimental error. It is usually denoted as H0.
Open Interval An Open Continuous Interval is one which excludes its endpoints, and is usually depicted with rounded brackets: (Minimum, Maximum).
Opportunity An Opportunity is an event or set of circumstances that may or may not occur, but if it does occur an Opportunity will have a beneficial effect on our plans, impacting positively on the cost, quality, schedule, scope compliance and/or reputation of our project or organisation.
Optimism Bias Optimism Bias is an expression of the inherent bias (often unintended) in an estimate output based on either incomplete or misunderstood input assumptions.
Outlier An outlier is a value that falls substantially outside the pattern of other data. The outlier may be representative of unintended atypical factors or may simply be a value which has a very low probability of occurrence.
Outturn Year Values ‘Outturn Year Values’ are values that have been adjusted to express an expectation of what might be incurred in the future due to escalation or other predicted changes. In some cases, these may be referred to as ‘Then Year Values’.
Paasche Index Paasche Indices are time-based indices which compare prices of commodities at a point in time with the equivalent prices for the Index Base Period, based on the quantities consumed at the current point in time in question
Parametric Estimating Method A Parametric Estimating Method is a systematic means of establishing and exploiting a pattern of behaviour between the variable that we want to estimate, and some other independent variable or set of variables or characteristics that have an influence on its value.
Payback Period The Payback Period is an expression of how long it takes for an investment opportunity to break even, i.e. to pay back the investment.
Pearson’s Linear Correlation Coefficient Pearson’s Linear Correlation Coefficient for two variables is a measure of the extent to which a change in the value of one variable can be associated with a change in the value of the other variable through a linear relationship. As such it is a measure of linear dependence or linearity between the two variables, and can be calculated by dividing the Covariance of the two variables by the Standard Deviation of each variable.
Peirce’s Criterion A test for multiple Outliers based on the deviation Z-Score of the suspect data point.
Percentile A Percentile is one of a hundred subsets from a set of ordered values which each nominally contain a hundredth of the total number of values in each subset. The term can also be used to express the values that divide the ordered values into the hundred ordered subsets.
Planned Value (PV) See Earned Value Management Abbreviations and Terminology.
Platykurtotic or Platykurtic An expression that the degree of Excess Kurtosis in a probability distribution is shallower than a Normal Distribution.
Power Function A Power Function of two variables is one in which the Logarithm of the dependent variable on the vertical axis produces a monotonic increasing or decreasing Straight Line when plotted against the Logarithm of the independent variable on the horizontal axis.
Precision (1) Precision is an expression of how close repeated trials or measurements are to each other.
(2) Precision is an expression of the level of exactness reported in a measurement, statistic or estimate.
Primary Data See Data Type.
Primary Driver See Estimate Drivers.
Probability Density Function (PDF) The Probability Density Function of a Continuous Random Variable expresses the rate of change in the probability distribution over the range of potential continuous values defined, and expresses the relative likelihood of getting one value in comparison with another.
Probability Mass Function (PMF) The Probability Mass Function of a Discrete Random Variable expresses the probability of the variable being equal to each specific value in the range of all potential discrete values defined. The sum of these probabilities over all possible values equals 100%.
Probability of Occurrence A Probability of Occurrence is a quantification of the likelihood that an associated Risk or Opportunity will occur with its consequential effects.
Quadratic Mean or Root Mean Square The Quadratic Mean of a set of n numerical data values is a statistic calculated by taking the square root of the Arithmetic Mean of the squares of the n values. As a consequence, it is often referred to as the Root Mean Square.
Quantile A Quantile is the generic term for a number of specific measures that divide a set of ordered values into a quantity of ranges with an equal proportion of the total number of values in each range. The term can also be used to express the values that divide the ordered values into such ranges.
Quantity-based Learning Curve A Quantity-based Learning Curve is an empirical relationship which reflects that the time, effort or cost to perform an activity reduces as the number of repetitions of that activity increases.
Quartile A Quartile is one of four subsets from a set of ordered values which nominally contain a quarter of the total number of values in each subset. The term can also be used to express the values that divide the ordered values into the four ordered subsets.
Queueing Theory Queueing Theory is that branch of Operation Research that studies the formation and management of queuing systems and waiting times.
Quintile A Quintile is one of five subsets from a set of ordered values which nominally contain a fifth of the total number of values in each subset. The term can also be used to express the values that divide the ordered values into the five ordered subsets.
Range The Range is the difference between the Maximum and Minimum observed values in a dataset, or the Maximum and Minimum theoretical values in a statistical distribution. In some circumstances, the term may be used to imply the difference between pessimistic and optimistic values from the range of potential values rather than an absolute range value.
Rate Metric A Rate is an Estimating Metric used to quantify how one variable’s value changes in relation to some measurable driver, attribute or parameter, and would be expressed in the form of a [Value] of one attribute per [Unit] of another attribute.
Ratio Metric A Ratio is an Estimating Metric used to quantify the relative size proportions between two different instances of the same driver, attribute or characteristic such as weight. It is typically used as an element of Estimating by Analogy or in the Normalisation of data.
Real Year Values ‘Real Year Values’ are values that have been adjusted to take account of historical or future inflationary effects or other changes, and are expressed in relation to the Current Year Values for any defined year. They are often referred to as ‘Constant Year Values’.
Regression Analysis Regression Analysis is a systematic procedure for establishing the Best Fit relationship of a predefined form between two or more variables, according to a set of Best Fit criteria.
Regression Confidence Interval The Regression Confidence Interval of a given probability is an expression of the Uncertainty Range around the Regression Line. For a known value of a single independent variable, or a known combination of values from multiple independent variables, the mean of all future values of the dependent variable will occur within the Confidence Interval with the probability specified.
Regression Prediction Interval A Regression Prediction Interval of a given probability is an expression of the Uncertainty Range around future values of the dependent variable based on the regression data available. For a known value of a single independent variable, or a known combination of values from multiple independent variables, the future value of the dependent variable will occur within the Prediction Interval with the probability specified.
Residual Risk Exposure The Residual Risk Exposure is the weighted value of the Risk, calculated by multiplying its Most Likely Value by the complement of its Probability of Occurrence (100% – Probability of Occurrence). It is used to highlight the relative value of the risk that is not covered by Risk Exposure calculation.
Risk A Risk is an event or set of circumstances that may or may not occur, but if it does occur a Risk will have a detrimental effect on our plans, impacting negatively on the cost, quality, schedule, scope compliance and/or reputation of our project or organisation.
Risk Exposure A Risk Exposure is the weighted value of the Risk, calculated by multiplying its Most Likely Value by its Probability of Occurrence.
See also Residual Risk Exposure.
Risk & Opportunity Ranking Factor A Risk & Opportunity Ranking Factor is the relative absolute exposure of a Risk or Opportunity in relation to all others, calculated by dividing the absolute value of the Risk Exposure by the sum of the absolute values of all such Risk Exposures.
Risk Uplift Factors A Top-down Approach to Risk Analysis may utilise Risk Uplift Factors to quantify the potential level of risk based on either known risk exposure for the type of work being undertaken based on historical records of similar projects, or based on a Subject Matter Expert’s Judgement.
R-Square (Regression) R-Square is a measure of the “Goodness of Fit” of a simple linear regression model to a set of data points. It is directly equivalent to the Coefficient of Determination that shows how much of the total variance in one variable can be explained by the variance in the other variable.
See also Adjusted R-Square.
Schedule Maturity Assessment (SMA) A Schedule Maturity Assessment provides a ‘health warning’ on the maturity of a schedule based on its underpinning assumptions and interdependencies, and takes account of the level of task definition available and historical evidence used.
Secondary Data See Data Type.
Secondary Driver See Estimate Drivers.
Skewness Coefficient The Fisher-Pearson Skewness Coefficient is an expression of the degree of asymmetry of a set of values around their Arithmetic Mean. A positive Skewness Coefficient indicates that the data has a longer tail on the right-hand side, in the direction of the positive axis; such data is said to be Right or Positively Skewed. A negative Skewness Coefficient indicates that the data has a longer tail on the left-hand side, in the direction of the negative axis; such data is said to be Left or Negatively Skewed. Data that is distributed symmetrically returns a Skewness Coefficient of zero.
Slipping and Sliding Technique A technique that compares and contrasts a Bottom-up Monte Carlo Simulation Cost evaluation of Risk, Opportunity and Uncertainty with a holistic Top-down Approach based on Schedule Risk Analysis and Uplift Factors.
Spearman’s Rank Correlation Coefficient Spearman’s Rank Correlation Coefficient for two variables is a measure of monotonicity of the ranks of the two variables, i.e. the degree to which the ranks move in the same or opposite directions consistently. As such it is a measure of linear or non-linear interdependence.
SPI (Schedule Performance Index – Cost Impact) See Earned Value Management Abbreviations and Terminology.
SPI(t) (Schedule Performance Index – Time Impact) See Earned Value Management Abbreviations and Terminology.
Spreadsheet Validation and Verification See individual terms: Validation and Verification.
Standard Deviation of a Population The Standard Deviation of an entire set (population) of data values is a measure of the extent to which the data is dispersed around its Arithmetic Mean. It is calculated as the square root of the Variance, which is the average of the squares of the deviations of each individual value from the Arithmetic Mean of all the values.
Standard Deviation of a Sample The Standard Deviation of a sample of data taken from the entire population is a measure of the extent to which the sample data is dispersed around its Arithmetic Mean. It is calculated as the square root of the Sample Variance, which is the sum of squares of the deviations of each individual value from the Arithmetic Mean of all the values divided by the degrees of freedom, which is one less than the number of data points in the sample.
Standard Error The Standard Error of a sample’s statistic is the Standard Deviation of the sample values of that statistic around the true population value of that statistic. It can be approximated by the dividing the Sample Standard Deviation by the square root of the sample size.
Stanford-B Unit Learning Curve A Stanford-B Unit Learning Curve is a variation of the Crawford Unit Learning Curve that allows for the benefits of prior learning to be expressed in terms of an adjustment to the effective number of cumulative units produced.
Statistics (1) The science or practice relating to the collection and interpretation of numerical and categorical data for the purposes of describing or inferring representative values of the whole data population from incomplete samples. (2) The numerical values, measures and context that have been generated as outputs from the above practice.
Stepwise Regression Stepwise Regression by Forward Selection is a procedure by which a Multi-Linear Regression is compiled from a list of independent candidate variables, commencing with the most statistically significant individual variable (from a Simple Linear Regression perspective) and progressively adding the next most significant independent variable, until such time that the addition of further candidate variables does not improve the fit of the model to the data in accordance with the accepted Measures of Goodness of Fit for the Regression.
Stepwise Regression by Backward Elimination is a procedure by which a Multi-Linear Regression is compiled commencing with all potential independent candidate variables and eliminating the least statistically significant variable progressively (one at a time) until such time that all remaining candidate variables are deemed to be statistically significant in accordance with the accepted Measures of Goodness of Fit.
Subject Matter Expert’s Opinion (Expert Judgement) Expert Judgement is a recognised term expressing the opinion of a Subject Matter Expert (SME).
SV (Schedule Variance – Cost Impact) See Earned Value Management Abbreviations and Terminology.
SV(t) (Schedule Variance – Time Impact) See Earned Value Management Abbreviations and Terminology.
Tertiary Data See Data Type.
Then Year Values ‘Then Year Values’ are values that have been adjusted to express an expectation of what might be incurred in the future due to escalation or other predicted changes. In some cases, these may be referred to as ‘Outturn Year Values’.
Three-Point Estimate See 3-Point Estimate.
Time Series Analysis Time Series Analysis is the procedure whereby a series of values obtained at successive time intervals is separated into its constituent elements that describe and calibrate a repeating pattern of behaviour over time in relation to an underlying trend.
An Additive/Subtractive Time Series Model is one in which the Predicted Value is a function of the forecast value attributable to the underlying Trend plus or minus adjustments for its relative Seasonal and Cyclical positions in time.
A Multiplicative Time Series Model is one in which the Predicted Value is a function of the forecast value attributable to the underlying Trend multiplied by appropriate Seasonal and Cyclical Factors.
Time-Based Learning Curve A Time-based Learning Curve is an empirical relationship which reflects that the time, effort or cost to produce an output from an activity decreases as the elapsed time since commencement of that activity increases.
Time-Constant Learning Curve A Time-Constant Learning Curve considers the output or yield per time period from an activity rather than the time or cost to produce a unit. The model assumes that the output increases due to learning, from an initial starting level, before flattening out asymptotically to a steady state level.
Time-Performance Learning Curve A Time-Performance Learning Curve is an empirical relationship that expresses the reduction in the average time or cost per unit produced per period as a power function of the cumulative number periods since production commenced.
Top-down Approach In a top-down approach to estimating, the estimator reviews the overall scope of work in order to identify the major elements of work and characteristics (drivers) that could be estimated separately from other elements. Typically, the estimator might consider a natural flow down through the Work Breakdown Structure (WBS), Product Breakdown Structure (PBS) or Service Breakdown Structure (SBS). The estimate scope may be broken down to different levels of WBS etc as required; it is not necessary to cover all elements of the task at the same level, but the overall project scope must be covered. The overall project estimate would be created by aggregating these high-level estimates. Lower level estimates can be created by subsequent iterations of the estimating process when more definition becomes available, and bridging back to the original estimate.
TRACEability A Basis of Estimate should satisfy the principles of TRACEability:
Transparent – clear and unambiguous with nothing hidden
Repeatable – allowing another estimator to reproduce the same results with the same information
Appropriate – it is justifiable and relevant in the context it is to be used
Credible – it is based on reality or a pragmatic reasoned argument that can be understood and is believable
Experientially-based – it can be underpinned by reference to recorded data (evidence) or prior confirmed experience
Transformation See Mathematical Transformation.
Trusted Source Estimating Method The Trusted Source Method of Estimating is one in which the Estimate Value is provided by a reputable, reliable or undisputed source. Typically, this might be used for low value cost elements. Where the cost element is for a more significant cost value, it would not be unreasonable to request the supporting Basis of Estimate, but this may not be forthcoming if the supporting technical information is considered to be proprietary in nature.
t-Test A t-Test is used for small sample sizes (< 30) to test probability of getting a sample’s test statistic (often the Mean), if the equivalent population statistic has an assumed different value. It is also used to test whether two samples could be drawn from the same population.
Tukey’s Fences A test for a single Outlier based on the Inter-Quartile Range of the data sample.
Type I Error A Type I Error is one in which we accept a hypothesis we should have rejected.
Type II Error A Type II Error is one in which we reject a hypothesis we should have accepted.
U-Test See Mann-Whitney U-Test.
Uncertainty Uncertainty is an expression of the lack of exactness around a variable, and is frequently quantified in terms of a range of potential values with an optimistic or lower end bound and a pessimistic or upper end bound.
Validation (Spreadsheet or Model) Validation is the process by which the assumptions and data used in a spreadsheet or model are checked for accuracy and appropriateness for their intended purpose.
See also Verification.
Variance of a Population The Variance of an entire set (population) of data values is a measure of the extent to which the data is dispersed around its Arithmetic Mean. It is calculated as the average of the squares of the deviations of each individual value from the Arithmetic Mean of all the values.
Variance of a Sample The Variance of a Sample of data taken from the entire population is a measure of the extent to which the sample data is dispersed around its Arithmetic Mean. It is calculated as the sum of squares of the deviations of each individual value from the Arithmetic Mean of all the values divided by the degrees of freedom, which is one less than the number of data points in the sample.
Verification (Spreadsheet or Model) Verification is the process by which the calculations and logic of a spreadsheet or model are checked for accuracy and appropriateness for their intended purpose.
See also Validation.
Wilcoxon-Mann-Whitney U-Test See Mann-Whitney U-Test.
Wright’s Cumulative Average Learning Curve Wright’s Cumulative Average Learning Curve is an empirical relationship that expresses the reduction in the cumulative average time or cost of each unit produced as a power function of the cumulative number units produced.
Z-Score A Z-Score is a statistic which standardises the measurement of the distance of a data point from the Population Mean by dividing by the Population Standard Deviation.
Z-Test A Z-Test is used for large sample sizes (< 30) to test probability of getting a sample’s test statistic (often the Mean), if the equivalent population statistic has an assumed different value.