D-optimality and \(D_ n\)-optimal designs for mixture regression models with logarithmic terms (Q1118301): Difference between revisions

Considered are the two logarithmic regression models \[ E(y)=\sum^{q}_{i=1}(\beta_ ix_ i+\gamma_ i \ln x_ i)\quad and\quad E(y)=\sum^{q}_{i=1}(\beta_ ix_ i+\gamma_ i \ln x_ i)+\sum^{q}_{i<j}\beta_{ij}x_ ix_ j \] over an experimental region with mixture- and boundary-restrictions. The authors present (without any theoretical analysis) the results of a computer search for D-optimal designs (exact and approximate) based on the Fedorov-Wynn algorithm.