Glossary Term | Glossary Definition |
---|---|
APT | Arbitrage Pricing Theory. |
assumption | An estimated value or input to a spreadsheet model. Assumptions capture the uncertainty of model data using probability distributions. |
bound | A maximum or minimum limit you set for each decision variable. |
certainty | The percentage of simulation results that fall within a range. |
coefficient of variability | A measure of relative variation that compares the standard deviation to the mean. Results can be represented in percentages for comparison purposes. |
constraint | A limitation that restricts the possible solutions to a model. You must define constraints in terms of decision variables. |
continuous | A variable that can be fractional (that is, it can take on any value between the lower and upper bounds). No step size is required and any given range contains an infinite number of possible values. Continuous also describes an optimization model that contains only continuous variables. |
correlation | A dependency that exists between assumption cells. |
correlation coefficient | A number between -1 and 1 that specifies mathematically the degree of positive or negative correlation between assumption cells. A correlation of 1 indicates a perfect positive correlation, minus 1 indicates a perfect negative correlation, and 0 indicates there is no correlation. |
decision variable | A variable in your model that you can control. |
deterministic | A model or system with no random variables that yields single-valued results. |
discrete variable | A variable that can only assume values equal to its lower bound plus a multiple of its step size; a step size is any number greater than zero, but less than the variable’s range. Discrete also describes an optimization model that contains only discrete variables. |
distribution | See probability distribution. |
efficient frontier | The curve that plots an objective value against changes to a requirement or constraint. A typical use is for comparing portfolio returns against different risk levels. |
efficient portfolio | Combinations of assets for which it is impossible to obtain higher returns without generating higher risk or lower risk without generating lower returns. An efficient portfolio lies directly on the efficient frontier. |
EOQ | Economic Order Quantity. |
feasible solution | A solution that satisfies any constraints imposed on the decision variables, as well as any requirements imposed on forecast statistics. |
final value | The last value that is calculated for a forecast during a simulation. The final value is useful for when a forecast contains a function that accumulates values across the trials of a simulation, or is a function that calculates the statistic of another forecast. |
forecast | A statistical summary of the mathematical combination of the assumptions in a spreadsheet model, output graphically or numerically. Forecasts are frequency distributions of possible results for the model. |
forecast objective | One forecast from a model that OptQuest uses as the primary goal of the optimization. OptQuest maximizes or minimizes a statistic of the forecast’s distribution. |
forecast statistic | Summary values of a forecast distribution, such as the mean, standard deviation, or variance. You control the optimization by maximizing or minimizing forecast statistics or setting them to a target value. |
frequency distribution | A chart that graphically summarizes a list of values by sub-dividing them into groups and displaying their frequency counts. |
heuristic | An approximate and self-educating technique for improving solutions. |
inventory | Any resource set aside for future use, such as raw materials, semifinished products, and finished products. Inventory also includes human, financial, and other resources. |
inventory level | The amount of inventory on hand, not counting ordered quantities not received. |
inventory position | The amount of inventory on hand plus any amount on order but not received, less any back orders. |
kurtosis | The measure of the degree of peakedness of a curve. The higher the kurtosis, the closer the points of the curve lie to the mode of the curve. A normal distribution curve has a kurtosis of 3. |
Latin hypercube sampling | A sampling method that divides an assumption’s probability distribution into intervals of equal probability. The number of intervals corresponds to the Sample Size option available in the Crystal Ball Run Preferences dialog. A random number is then generated for each interval. Compared with conventional Monte Carlo sampling, Latin hypercube sampling is more precise because the entire range of the distribution is sampled in a more even, consistent manner. The increased accuracy of this method comes at the expense of added memory requirements to hold the full Latin hypercube sample for each assumption. |
linear | A mathematical relationship where all terms in the formulas can only contain a single variable multiplied by a constant. For example, 3x - 1.2y is a linear relationship since both the first and second term involve only a constant multiplied by a variable. |
maximum | The largest value in a dataset. |
mean | The familiar arithmetic average of a set of numerical observations: the sum of the observations divided by the number of observations. |
mean standard error | The standard deviation of the distribution of possible sample means. This statistic gives one indication of how accurate the simulation is. |
median | The value midway (in terms of order) between the smallest possible value and the largest possible value. |
metaheuristic | A family of optimization approaches that includes genetic algorithms, simulated annealing, tabu search, scatter search, and their hybrids. |
minimum | The smallest value in a dataset. |
mixed | A type of optimization model that has both discrete and continuous decision variables. |
mode | The value that, if it exists, occurs most often in a data set. |
model | A representation of a problem or system in a spreadsheet application such as Excel. |
multiobjective optimization | A technique that combines multiple, often conflicting objectives, such as maximizing returns and minimizing risks, into one objective. |
nonlinear | A mathematical relationship where one or more terms in the formulas are nonlinear. Terms such as x2, xy, 1/x, or 3.1x make nonlinear relationships. See linear. |
NPV | Net Present Value. The NPV equals the present value minus the initial investment. |
objective | A forecast formula in terms of decision variables that gives a mathematical representation of the model’s goal. |
optimal solution | The set of decision variable values that achieves the best outcome. |
optimization | A process that finds the optimal solution to a model. |
optimization model | A model that seeks to maximize or minimize some quantity (the objective), such as profit or risk. |
order quantity | The standard amount of product you reorder when inventory reaches the reorder point. |
percentile | A number on a scale of zero to one hundred that indicates the percent of a probability distribution that is equal to or below a value (default definition). |
performance | For an optimization program, the ability to find high-quality solutions as fast as possible. |
probability | The likelihood of an event. |
probability distribution | A set of all possible events and their associated probabilities. |
random number | A mathematically selected value which is generated (by a formula or selected from a table) to conform to a probability distribution. |
random number generator | A method implemented in a computer program that is capable of producing a series of independent, random numbers. |
range | The difference between the largest and smallest values in a data set. |
rank correlation | A method whereby Crystal Ball replaces assumption values with their ranking from lowest value to highest value (1 to N) prior to computing the correlation coefficient. This method lets you ignore the distribution types when correlating assumptions. |
RAROC | A multiobjective function that calculates the Risk-adjusted Return On Capital. |
reorder point | The inventory position when you reorder. |
requirement | A restriction on a forecast statistic that requires the statistic to fall between specified lower and upper limits for a solution to be considered feasible. |
risk | The uncertainty or variability in the outcome of some event or decision. |
risk factor | A number representing the riskiness of an investment relative to a standard, such as U.S. Treasury bonds, used especially in APT. |
safety stock | The additional quantity kept in inventory above planned usage rates. |
seed value | The first number in a sequence of random numbers. A given seed value produces the same sequence of random numbers for assumption values every time you run a simulation. |
sensitivity | The amount of uncertainty in a forecast cell that is a result of both the uncertainty (probability distribution) and model sensitivity of an assumption or decision variable cell. |
sensitivity analysis | The computation of a forecast cell’s sensitivity with respect to the assumption or decision variable cells. |
simulation | A set of Crystal Ball trials. OptQuest finds optimal solutions by running multiple simulations for different sets of decision variable values. |
skewed | An asymmetrical distribution. |
skewness | The measure of the degree of deviation of a curve from the norm of an asymmetric distribution. The greater the degree of skewness, the more points of the curve lie on one side of the peak of the curve as compared to the other side. A normal distribution curve, having no skewness, is symmetrical. |
spreadsheet model | Any spreadsheet that represents an actual or hypothetical system or set of relationships. |
standard deviation | The square root of the variance for a distribution. A measurement of the variability of a distribution, that is, the dispersion of values around the mean. |
step size | Defines the difference between successive values of a discrete decision variable in the defined range. For example, a discrete decision variable with a range of 1 to 5 and a step size of 1 can take on only the values 1, 2, 3, 4, or 5; a discrete decision variable with a range of 0 to 17 with a step size of 5 can take on only the values 0, 5, 10, and 15. |
stochastic | A model or system with one or more random variables. |
STOIIP | Stock Tank Oil Initially In Place. STOIIP is the estimated reserves of an oil field in millions of barrels (mmbbls). |
trial | A three-step process in which Crystal Ball generates random numbers for assumption cells, recalculates the spreadsheet models, and displays the results in a forecast chart. A Crystal Ball simulation is made up of multiple trials. |
variable | A quantity that might assume any one of a set of values and is usually referenced by a formula. |
variance | The square of the standard deviation, where standard deviation is approximately the average of the sum of the squares of the deviations of a number of observations (n) from their mean value (except the sum is divided by n-1 instead of n, which would yield a true average). Variance can also be defined as a measure of the dispersion, or spread, of a set of values about a mean. When values are close to the mean, the variance is small. When values are widely scattered about the mean, the variance is larger. |
wizard | A feature that leads you through the steps to create and run an optimization model. This wizard presents panels for you to complete in the proper order. |