Facebook Instagram Twitter RSS Feed PodBean Back to top on side

A comparison of optimum design algorithms for regressions with correlated observations — a computational study

In: Tatra Mountains Mathematical Publications, vol. 17, no. 3
Ernst Glatzer - Werner G. Müller
Detaily:
Rok, strany: 1999, 149 - 156
O článku:
There exists several algorithms for the calculation of optimum (or improved) designs in the regression case under correlated observations with a given covariance function. One of them was proposed by Brimkulov et al. (1980) [U. N. Brimkulov, G. K. Krug, V. L. Savanov: Numerical construction of exact experimental designs when the measurements are correlated, Zavodskaya Laboratoria 36 (1980), 435–442. (In Russian.)] and it can be interpreted as replacing expressions in the definition of classical (uncorrelated observation) algorithms by their analogous versions in the correlated setup. Unfortunately convergence of such an algorithm to the true optimum is not guaranteed and this will be shown in several examples. A comparison with an alternative approach, the recently proposed algorithm by Müller and Pázman (1998) [W. G. Müller, A. Pázman: An algorithm for the computation of optimum designs under a given covariance structure, Comput. Statistics 14 (1999)], will be given.
Ako citovať:
ISO 690:
Glatzer, E., Müller, W. 1999. A comparison of optimum design algorithms for regressions with correlated observations — a computational study. In Tatra Mountains Mathematical Publications, vol. 17, no.3, pp. 149-156. 1210-3195.

APA:
Glatzer, E., Müller, W. (1999). A comparison of optimum design algorithms for regressions with correlated observations — a computational study. Tatra Mountains Mathematical Publications, 17(3), 149-156. 1210-3195.