We discuss methodology for multidimensional scaling (MDS) and its implementation in two software systems, GGvis and XGvis. MDS is a visualization technique for proximity data, that is, data in the form of N × N dissimilarity matrices. MDS constructs maps ("configurations," "embeddings") in IR k by interpreting the dissimilarities as distances. Two frequent sources of dissimilarities are high-dimensional data and graphs. When the dissimilarities are distances between high-dimensional objects, MDS acts as a (often nonlinear) dimension-reduction technique. When the dissimilarities are shortest-path distances in a graph, MDS acts as a graph layout technique. MDS has found recent attention in machine learning motivated by image databases ("Isomap"). MDS is also of interest in view of the popularity of "kernelizing" approaches inspired by Support Vector Machines (SVMs; "kernel PCA").This article discusses the following general topics: (1) the stability and multiplicity of MDS solutions; (2) the analysis of structure within and between subsets of objects with missing value schemes in dissimilarity matrices; (3) gradient descent for optimizing general MDS loss functions ("Strain" and "Stress"); (4) a unification of classical (Strain-based) and distance (Stress-based) MDS.Particular topics include the following: (1) blending of automatic optimization with interactive displacement of configuration points to assist in the search for global optima; (2) forming groups of objects with interactive brushing to create patterned missing values in MDS loss functions; (3) optimizing MDS loss functions for large numbers of objects relative to a small set of anchor points ("external unfolding"); and (4) 445We show applications to the mapping of computer usage data, to the dimension reduction of marketing segmentation data, to the layout of mathematical graphs and social networks, and finally to the spatial reconstruction of molecules.

We propose to furnish visual statistical methods with an inferential framework and protocol, modelled on confirmatory statistical testing. In this framework, plots take on the role of test statistics, and human cognition the role of statistical tests. Statistical significance of 'discoveries' is measured by having the human viewer compare the plot of the real dataset with collections of plots of simulated datasets. A simple but rigorous protocol that provides inferential validity is modelled after the 'lineup' popular from criminal legal procedures. Another protocol modelled after the 'Rorschach' inkblot test, well known from (pop-)psychology, will help analysts acclimatize to random variability before being exposed to the plot of the real data. The proposed protocols will be useful for exploratory data analysis, with reference datasets simulated by using a null assumption that structure is absent. The framework is also useful for model diagnostics in which case reference datasets are simulated from the model in question. This latter point follows up on previous proposals. Adopting the protocols will mean an adjustment in working procedures for data analysts, adding more rigour, and teachers might find that incorporating these protocols into the curriculum improves their students' statistical thinking.