munfold

An R package which provides Schoenemann’s algorithm for metric multidimensional unfolding and Procrustes rotation of unfolding results.

Multi-dimensional unfolding is a procedure to recover positions of two sets of points from a matrix of distances between the points of both groups. For some of my research work, I implemented the algorithm for metric multi-dimensional unfolding that Peter Schönemann published back in 1970 in Psychometrika. This is not the most advanced algorithm, however it is quite robust and quick when a large number of points is involved. Note that it is an algorithm for metric unfolding. That is, the input data need to be interpretable as real distances.

The package that contains my implementation of Schönemann’s algorithm is called munfold. It is now published on CRAN. Development occurs on GitHub, where both releases and the development tree can be found.

A demonstration of multidimensional unfolding. The left panel shows the original configuration, the right panel shows the reconstructed configuration. Note that the Schoenemann algorithm leads to a successful reconstruction of the circle only if the positions of the points are very slightly perturbed.