Abstract
Multidimensional scaling: recovery of metric information wit incomplete designs. The objective of this investigation was to prove the efficiency of three incomplete designs which can be used to select the data entry in MultiDimensional Scaling when the number of stimuli is high. Three types of designs were compared: a ciclic design (Spence & Domoney, 1974), two aleatory designs and a design with the biggest dissimilarities. When the minimum percentage of necessary information proposed by Spence and Domoney was used, all of the designs showed a similar degree of efficiency. However, with lower levels of information the ciclic design showed the worst results, while the other two types of designs mantained their efficiency, being the aleatory designs the ones which allowed to obtain satisfactory solutions even with the lowest percentages of entry information.
