Prof. Dr. H. Hebbel
Empirische Wirtschaftsforschung und Datenverarbeitung


Louvain

Next: Achievement by desirability Up: Statistical aspects of Multi-Criteria-Optimisation Previous: The Concept of Desirability

What is a must for a desirability function d?

Step 1+2: Target value and maximisation problems can be formulated.
Parameters useful for target value problems are:

d monotonically increasing in $(- \infty, T)$ and decreasing in $(T, \infty)$.

Guarantees pareto-optimality of any desirability-optimum!

Nice to have: weights $\beta_l, \beta_r$ for deviations to the left or the right of T.

Step 3: Use some kind of mean value.

Most popular choice is the geometric mean.

Important feature: D(X) = 0, if any di(X) = 0

"If one of the properties is completely unacceptable, the product as a whole is unacceptable."

$
DI(P) := \sqrt[Z]{\prod_1^Z d_i(Y_i)}
= \sqrt[Z]{\prod_1^Z d_i(f_i(X_1,X_2, \ldots, X_F,\epsilon_i))}
$

Z-th root is needed for comparability of single and overall results.

Alternative: min D

\begin{displaymath}D(P) := \max_{{\bf X}} \min_{i = 1, \ldots, Z} d_i({\bf X}). \end{displaymath}

"A product is only as good as its worst property."

No alternative, but often used in practice: arithmetical mean.
This allows partly unusable products.


Dipl.-Stat. Detlef Steuer
1999-10-04


E-Mail-Kontakt steuer@unibw-hamburg.de | Druckdatum: 01.06.2004 - 18:39:55