Conclusion

Desirability Indices are very successful in practical applications.
Consequences of noisy data are understimated in most existing literature.
Ongoing working on analytical determination of prognosis-intervals for X_opt
Simulation studies together with $(\text{leave-one-out})^2$ are an approach to identify badly conditioned situations and keep the indices a well behaving black box.

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