06.07.2013 Views

Logical Decisions - Classweb

Logical Decisions - Classweb

Logical Decisions - Classweb

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

no intrinsically correct shape for a SUF, the assessment process<br />

must be subjective. There are two commonly used methods for<br />

assessing a SUF -- mid-level splitting and assessments using<br />

probabilities. These approaches have different theoretical<br />

foundations that are mostly of academic interest. If you know the<br />

differences, you will know which technique you will want to use.<br />

If not, use the technique you feel most comfortable with.<br />

SUF Assessment using Mid-Level Splitting. The mid-level<br />

splitting procedure seeks to identify the level that is exactly midway<br />

(half-way) in preference between a low level and a high level<br />

for a measure. The SUF::Assess Value option implements the<br />

mid-level splitting procedure in LDW. Remember from the<br />

discussion on page 9-15 that this mid-preference level could well<br />

be different from the average of the two ends of the range.<br />

You define the mid-preference level by asking the decision maker<br />

questions about changes in the measure. Specifically the midpreference<br />

level (call it L) is where you prefer a change from the<br />

least preferred level (call it L0) to L and a change from L to the<br />

most preferred level (call it L1) equally. In other words, you want<br />

to find the level L that divides the range from L0 to L1 into two<br />

pieces, each representing an equal change in utility.<br />

In more mathematical terms, if U(x) represents the SUF for the<br />

measure, we want to find the point where<br />

U(L) - U(L0) = U(L1) - U(L).<br />

Remember that you should always prefer L to L0 and you should<br />

always prefer L1 to L. However, you still may feel that the change<br />

from L0 to L is more significant or important than the change from<br />

L to L1. You may feel this even though the result of the second<br />

change (L1) is the most preferred level overall.<br />

Assume when asking these questions about one measure that the<br />

levels of all of the other measures remain constant. You should<br />

think of the questions of in terms of simplified hypothetical<br />

alternatives, even if the levels in the questions are similar to those<br />

for the real alternatives.<br />

If the decision maker cannot answer the questions without<br />

knowing a specific level of another measure, there may be a<br />

preferential dependence between the two measures. Modeling<br />

preference dependencies is beyond the current abilities of LDW.<br />

9-16 Section 9 -- In Depth

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!