Clustering Design (Initial Version)

 

 

 

 

 

The Cluster Design Method was investigated as an alternative to generate designs in our group. The strategy was as follows in its initial form: (i) generate a full factorial design as an initial enumeration; (ii) add a column with uniformly distributed random numbers to the full factorial design; (iii) generate k clusters with the k-means algorithm, with k being the number of necessary regression coefficients plus one;(iv) retrieve the k-medoids associated to the k clusters; (v) delete the values associated to the column with the random numbers, and (vi) present the experimental design.