Logical Decisions - Classweb
Logical Decisions - Classweb Logical Decisions - Classweb
Tradeoff Trial program decides which formula to use based on your input in the Assess SUFs option. The linear SUF formula is just the formula that gives a straight line from the utility of the least preferred level of the sub-range to the utility of the most preferred level of the sub-range. The formula for a linear SUF is: U(X) = a + bX, where a and b are computed scaling constants and X is a level for the measure. The exponential SUF formula is used to fit a smooth curve to three points -- the least preferred level of the sub-range, the most preferred level of the sub-range and the mid-preference level of the sub range. The formula is (-cX) U(X) = a + (be ), where a, b, and c are computed scaling constants and e is the mathematical constant 2.718... See also: Assess SUFs option, Level, Measure, SUF, Utility. A tradeoff is a pair of equally preferred hypothetical alternatives that differ on only two measures. Alternative B has a more preferred level on measure 1 and a less preferred level on measure 2 while alternative A has a less preferred level on measure 1 and a more preferred level for measure 2. The levels of the measures are set so that a change in measure 1 just compensates for a corresponding change in measure 2. Equally preferred alternatives should have equal overall utilities, and since alternatives A and B differ only in measures 1 and 2, these compensating changes can be used to compute the relative weights for measures 1 and 2. See also: Alternative, Level, Measure, Utility, Weight. A trial is a single iteration in a Monte Carlo simulation. In LDW a trial results in the evaluation of the utility of an alternative based on a possible resolution of its uncertainties. In the trial, each probabilistic level is replaced by a deterministic sample from its probability distribution. The samples are generated using a random 12-10 Section 12 -- Glossary
Utility Weight number generator and the inverse probability distribution for the probabilistic level. Once all the probabilistic levels have been replaced, LDW computes and saves the requested utility. After many trials have been computed, the cumulative probability distribution on the alternative's utility can be estimated. See also: Alternative, Monte Carlo Simulation, Probabilistic Level, Sample, Utility. Utility is a standardized measure of the relative desirability of a given level or set of levels for an alternative. Utilities are the output of a Multi-measure Utility Function (MUF) or Single-measure Utility Function (SUF). They are used to convert the levels for measures, which are based on scales with potentially very different units, into a comparable scale with a range defined to go from 0.0 to 1.0. Utility functions generally assign a utility of 0.0 to the least preferred level for a measure, and assign 1.0 to the most preferable level for a measure. Alternatives with utilities closer to 1.0 are preferred. See also: Alternative, Level, Measure, MUF, SUF. Weights are a casual term for the scaling constants (small ks) associated with the members of a goal in the MUF for a goal. Weights have no intrinsic importance, but do provide an indication of the relative importance of the measures given the ranges found for a set of alternatives. The weights in a MUF are determined by the tradeoffs that define the MUF. The tradeoffs define a unique set of weights that will allow all of the equally preferred alternatives in the tradeoffs to get the same overall utility. See also: Alternative, Goal, Measure, MUF, Range, Small k, Tradeoff, Utility. Section 12 -- Glossary 12-11
- Page 311 and 312: Interpreting the Ranking Results LD
- Page 313: S E C T I O N Examples 10
- Page 316 and 317: the idea that other manufactures an
- Page 318 and 319: The completed goals hierarchy is sh
- Page 320 and 321: Buying a House The ranking results
- Page 322 and 323: Figure 10-5. Goals hierarchy for bu
- Page 324 and 325: Overall goal Quality goal Costs goa
- Page 326 and 327: The preference assessments were don
- Page 328 and 329: Figure 10-9. Goals hierarchy for re
- Page 330 and 331: ! Noise, ! Agricultural Impacts, an
- Page 333: S E C T I O N Commands Summary 11
- Page 336 and 337: Assess Menu Edit Menu The Assess me
- Page 338 and 339: File Menu Edit::Insert Lets you add
- Page 340 and 341: utility function” if the active g
- Page 342 and 343: Matrix Menu ! Hierarchy -- options
- Page 344 and 345: Results::Dynamic Sensitivity See th
- Page 346 and 347: Review::Compute Utilities Compute t
- Page 348 and 349: Tradeoff Menu View Menu LDW display
- Page 351: S E C T I O N Glossary 12
- Page 354 and 355: Common Units Certainty Equivalent C
- Page 356 and 357: LDW File Level Lottery Measure assi
- Page 358 and 359: MUF MUF Formula See also: Alternati
- Page 360 and 361: Preference Set Probabilistic Level
- Page 365: Bibliography B
- Page 368 and 369: The classic reference on multiple m
- Page 371 and 372: Appendix This appendix describes th
- Page 373: = .125 Thus the weight for cost is
- Page 377 and 378: Index adjusted AHP ................
- Page 379: computing ................8 - 3 int
Tradeoff<br />
Trial<br />
program decides which formula to use based on your input in the<br />
Assess SUFs option.<br />
The linear SUF formula is just the formula that gives a straight line<br />
from the utility of the least preferred level of the sub-range to the<br />
utility of the most preferred level of the sub-range. The formula for a<br />
linear SUF is:<br />
U(X) = a + bX,<br />
where a and b are computed scaling constants and X is a level<br />
for the measure.<br />
The exponential SUF formula is used to fit a smooth curve to three<br />
points -- the least preferred level of the sub-range, the most preferred<br />
level of the sub-range and the mid-preference level of the sub range.<br />
The formula is<br />
(-cX)<br />
U(X) = a + (be ),<br />
where a, b, and c are computed scaling constants and e is the<br />
mathematical constant 2.718...<br />
See also: Assess SUFs option, Level, Measure, SUF, Utility.<br />
A tradeoff is a pair of equally preferred hypothetical alternatives that<br />
differ on only two measures. Alternative B has a more preferred<br />
level on measure 1 and a less preferred level on measure 2 while<br />
alternative A has a less preferred level on measure 1 and a more<br />
preferred level for measure 2. The levels of the measures are set so<br />
that a change in measure 1 just compensates for a corresponding<br />
change in measure 2. Equally preferred alternatives should have<br />
equal overall utilities, and since alternatives A and B differ only in<br />
measures 1 and 2, these compensating changes can be used to<br />
compute the relative weights for measures 1 and 2.<br />
See also: Alternative, Level, Measure, Utility, Weight.<br />
A trial is a single iteration in a Monte Carlo simulation. In LDW a<br />
trial results in the evaluation of the utility of an alternative based on<br />
a possible resolution of its uncertainties. In the trial, each<br />
probabilistic level is replaced by a deterministic sample from its<br />
probability distribution. The samples are generated using a random<br />
12-10 Section 12 -- Glossary