Data Quality, Level of Aggregation, and Nonmetric Multidimensional Scaling
George S. Day, Terry H. Deutscher, Adrian B. Ryans · 2016
concern should be with the metric determinacy of the obtained configuration, not with stress. An index of metric determinacy (M) is the square of the correlation of the distances from the true configuration with those from the recovered configuration. Sherman [12] investigated the relationship between V 1 - M and stress. He concluded that stress tends to be related linearly to metric determinacy when the number of points is large in relation to the number of recovered dimensions and when the level of error in the data is low. Thus under these conditions when stress is low, the recovered configuration is probably very similar to the underlying true configuration.