Skip to main content
Log in

Versuchspläne zur stochastischen Optimierung

Experimental designs for stochastic optimization

  • Published:
Computing Aims and scope Submit manuscript

Zusammenfassung

In der experimentell medizinischen Forschung mangelt es an Plänen zur Auffindung optimaler Stufenkombinationen von zwei oder mehr Faktoren. Dazu werdenk-dimensionale orthogonale zentrale zusammengesetzte Versuchspläne 2. Ordnung, bestehend aus einem vollständigen 2k-Plan oder einer geeigneten fraktionellen Replikation davon, einem oder zwei Axialplänen undz Versuchen im Mittelpunkt (z=0, 1, 2, ...) angeführt. Die Pläne unterscheiden sich hauptsächlich in der Genauigkeit, mit der die Regressionskoeffizienten, insbesondere die quadratischen, geschätzt werden, und im erforderlichen Versuchsbereich. Fallen zwei Axialpläne zusammen, so wird die vergrößerte Versuchsanzahl durch die Möglichkeit einer Adäquatheitsprüfung des Modells aufgewogen. Diese Pläne wurden in Anlehnung an die von Box und Wilson vor allem für Optimierungsprobleme in der chemischen Forschung angegebenen Prozeduren entwickelt. Falls erforderlich, können Begleitvariable durch entsprechende Kovarianzanalysen berücksichtigt werden.

Summary

In medical research there is a lack of experimental designs aiming at an optimal stage-combination of two or more factors. For this purposek-dimensional orthogonal central composite designs were constructed. They consist of a factorial 2k-design or of a suitable fractional replication, of one or two axial plans and ofz experiments in the center (z=0, 1, 2, ...). The designs differ in their precision of estimating the regression coefficients and, concomitantly, in the experimental region covered by them. If two axial plans coincide, the increase in the number of experiments is compensated by the possibility of testing the model for adequacy. These designs were based on procedures of Box and Wilson, which were primarily developed for optimization problems in chemical research. Inclusion of concomitant variables, if necessary, is provided for.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Literatur

  1. Blum, J. R.: Multidimensional stochastic approximation methods. Ann. Math. Stat.25, 737–744 (1954).

    Google Scholar 

  2. Box, G. E. P.: The exploration and exploitation of response surfaces. Some general considerations and examples. Biometrics10, 16–60 (1954).

    Google Scholar 

  3. Box, G. E. P.: Evolutionary operation: A method for increasing industrial productivity. Appl. Stat.6, 81–101 (1957).

    Google Scholar 

  4. Box, G. E. P., andJ. S. Hunter: A confidence region for the solution of a set of simultaneous equations with an application to experimental design. Biometrika41, 190–199 (1954).

    Google Scholar 

  5. Box, G. E. P., andJ. S. Hunter: Multifactor experimental designs for exploring response surfaces. Ann. Math. Stat.28, 195–241 (1957).

    Google Scholar 

  6. Box, G. E. P., andJ. S. Hunter: Condensed calculations for evolutionary operations. Technometrics1, 77–95 (1959).

    Google Scholar 

  7. Box, G. E. P., andK. B. Wilson: On the experimental attainment of optimum condition. J. Roy. Stat. Soc.B 13, 1–45 (1951).

    Google Scholar 

  8. Box, G. E. P., andP. V. Youle: The exploration and exploitation of response surfaces: An example of the link between the fitted surface and the basic mechanism of the system. Biometrics11, 287–323 (1955).

    Google Scholar 

  9. Brooks, S. H., andM. R. Mickey: Optimum estimation of gradient direction in steepest ascent experiments. Biometrics17, 48–56 (1961).

    Google Scholar 

  10. Collatz, L.: Funktionalanalysis und numerische Mathematik. Berlin-Göttingen-Heidelberg-New York: Springer 1964.

    Google Scholar 

  11. Davies, O. L. (ed.): The design and analysis of industrial experiments. 2. Aufl. London: Oliver and Boyd. 1967.

    Google Scholar 

  12. Hicks, Ch. R.: Fundamental concepts in the designs of experiments. New York: Holt, Rinehart and Winston. 1964.

    Google Scholar 

  13. Hunter, W. G., andJ. R. Kittrel: Evolutionary operation: A review. Technometrics8, 389–397 (1966).

    Google Scholar 

  14. Kesten, E.: Accelerated stochastic approximation. Ann. Math. Stat.29, 41–59 (1958).

    Google Scholar 

  15. Kiefer, J., andJ. Wolfowitz: Stochastic estimation of the maximum of a regression function. Ann. Math. Stat.23, 462–466 (1952).

    Google Scholar 

  16. Lapidus, L., E. Shapiro, S. Shapiro, andR. E. Stillman: Optimization of process performance. A. I. Ch. E. J.7, 288–294 (1961).

    Google Scholar 

  17. Linder, A.: Planen und Auswerten von Versuchen, 3. Aufl. Basel: Birkhäuser. 1969.

    Google Scholar 

  18. Messikommer, B. H.: Die Anwendung der Box-Wilson-Methode in der chemischen Industrie. Unternehmensforschung4, 112–137 (1960).

    Google Scholar 

  19. Sacks, J.: Asymptotic distribution of stochastic approximation procedures. Ann. Math. Stat.29, 373–405 (1958).

    Google Scholar 

  20. Scheiber, V.: Eine Anwendung des Optimierungsverfahrens von Box und Wilson in der experimentellen Medizin. Int. J. clin. Pharmacol.5, 484–498 (1972).

    Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Additional information

Mit 5 Abbildungen

Rights and permissions

Reprints and permissions

About this article

Cite this article

Scheiber, V. Versuchspläne zur stochastischen Optimierung. Computing 9, 383–399 (1972). https://doi.org/10.1007/BF02241611

Download citation

  • Received:

  • Issue Date:

  • DOI: https://doi.org/10.1007/BF02241611

Navigation