MDS can be used with any proximity matrix in ClustanGraphics. You can read in a matrix of observed proximities, or compute one from your data. You can also carry out a cluster analysis on a large amount of data, truncate your cluster model to a manageable number of clusters, with or without outliers, and then obtain a MDS configuration that represents graphically your final cluster model.
Examples of proximities that can be read are:
Similarities of pairs of entities, for example judgements on a scale from "0" meaning maximum dissimilarity to "9" meaning maximum similarity.
Dominance of one entity over another, meaning that one entity is larger than, or preferred to, or having more of some desired attribute than another entity. This type of data can occur in focus groups and marketing applications.
Comparisons between specimens in DNA-analysis, such as the number of proteins that two species have in common, or simple matching coefficients between pairs of DNA arrays such as are illustrated by theProteins and Primates examples in ClustanGraphics.
Communications between entities, such as transactions between the nodes of an intranet, interactions between organizational units, or traffic flows between cities.
If the data matrix that is input to ClustanGraphics containmissing values, ClustanMDS can be used to find a configuration in several dimensions that fits the corresponding distance matrix and is complete. It therefore offers an alterative way of imputing missing values.