Glossary

This glossary of terms used in TMIP-EMAT is based on, and heavily influenced by, the robust decision making glossary published by The RAND Corporation. There are some minor difference and additions that are specifically tailored to the TMIP-EMAT approach demonstrated here.

box
A box is a subset of cases in a design of experiments, containing only those cases that meet certain restrictions. Typically, these restrictions are expressed as limited ranges on a particular set of parameters or performance measures. A box can be found as the result of certain EMA methodologies (e.g. PRIM, CART) or simply generated manually during an exploration of the data.
CART
Classification and Regression Trees is a relatively greedy but long-standing approach to developing interesting boxes for model exploration. (Breiman, 1984).
case
A case is a single modeling experiment, defined by a particular policy, scenario, and set of performance measures outcomes.
core model
A model that represents some relationships between exogenous uncertainties, policy levers, and performance measures. The model can be implemented as a Python function, an Excel workbook, or any other computer-based model that can be evaluated automatically from Python (e.g., at the command line).
design
A design of experiments is a group of cases generated in some systematic manner. Designs can be as simple as a Monte Carlo sample, where a number of independent random draws are made from the possible input parameters, or more structured random (e.g. latin hypercube) or non-random (e.g. factorial) designs
experiment
The execution of the model to calculate performance measures. See also case.
lever
The L in XLRM, a lever is a single policy strategy that might be implemented by decision maker(s). Like exogenous uncertainties, a lever is an input to the underlying model. Unlike exogenous uncertainties, a lever does not have any distributional assumptions.
measure
The M in XLRM, this is a single performance measure that can be used to evaluate the performance of the system, including whether or not decision maker’s goals are being met. The current version of TMIP-EMAT assumes that each measure is a single scalar outcome, although any number of performance measures can be included, so that a vector of outcomes (e.g., traffic volumes across a sequence of screen lines along a corridor) can be jointly considered. Future version of TMIP-EMAT may incorporate time series and/or array-based performace measures, which are explicitly available in the underlying EMA Workbench source code.
meta-model
A meta-model is an analytical approximation of some core model. Typically a meta-model is used when the execution of the core model is computationally expensive. The meta-model complements the core model and can approximate the results of the core model in a fraction of the time, but it is not a replacement for having a core model in the first place.
model
A model that represents some relationships between exogenous uncertainties, policy levers, and performance measures. The model can be a core model used directly, or a meta-model derived from a core model.
parameter
A parameter is an input to a model. This is a generic term that groups together exogenous uncertainties, policy levers, as well as any other constant values that are passed as inputs to a model (as may happen for model inputs that are potentially changeable but are set to fixed values within the current exploratory scope).
policy
A set of values for all the various policy levers defined in an exploratory modeling scope.
PRIM
The Patient Rule Induction Method, a “bump hunting” technique introduced by Friedman and Fisher (1999).
scenario
A set of values for all the various exogenous uncertainties defined in an exploratory modeling scope.
scope
An exploratory modeling scope is a collection of definitions for the exogenous uncertainties, policy levers, and performance measures under consideration.
SPLOM
The Scatter Plot Matrix, or SPLOM, is a visualization method. It is a collection of two-dimensional scatter plots arranged in a matrix, where each column of plots shares a common x-axis defintion, and each row shares a common y-axis definition.
uncertainty
The X in XLRM, and uncertainty is a single exogenous risk factor that represents an unknown and uncontrolled future state of the system under study. An uncertainty is an input to the underlying model, and is characterized by a random variable distribution. In many exploratory modeling contexts and for certain types of exploratory modeling analysis, analysts explicitly disavow making probabilistic statements about the possible values of the uncertainty factors, and instead merely characterize uncertainty as a range (for numeric-type uncertainties) or a set of possible values (for boolean or categorical type uncertainties).
XLRM

A general framework for conducting exploratory analysis, as proposed in Lempert et al (2003). The letters refer two four principal components of the analysis: