| NEWS | R Documentation |
Recoding of transformations:
Ordinal transformations are coded in C++ and therefore much faster.
Added monotone spline transformations (type = mspline with number of interior knots spline.intKnots and degree of the spline spline.degree = 2.
smacofConstraint allows for transformations of the external variables through constraint.type ("ratio", "interval", "ordinal", "spline", "spline", or "mspline"). Missing values (NA) are estimated by the model. Choosing the number of external variables equal to the number of dimensions together with constraint.type = "ordinal", constraint.ties = "primary" will effectively restrict the configuration to parallel regions defined by the categories of the external variables.
Stress values reported by verbose = TRUE during iterations of smacofSym, smacofConstraint, and smacofRect are normalized Stress values.
Gravity model for dissimilarity computation based on co-occurrences added: see gravity().
The circle argument in smacofRect() allows for circular restrictions on the configurations.
All smacof functions now return Stress-1 (before it was squared Stress-1); the value return by the function is now called stress.
The metric argument is deprecated. Now type should be used (either "ordinal", "interval", or "ratio")
for pretty much all smacof functions.
smacofSphere.primal() and smacofSphere.dual() are now merged into a single function smacofSphere() with a corresponding algorithm argument.
smacofConstraint() now returns weight matrix C.
Plot updates:
For unfolding (smacofRect()) configurations: joint = TRUE as default.
Configuration plot looks a bit fancier now (labels and points as default), also aspect ratio = 1 as default.
A note in plot help file added that in the stress plot the SPP from a various smacof functions is rescaled to percentages (which sum up to 100).
Datasets:
Labels for rectangle dataset now provided in English.
Dataset on Republican statements added (document-term matrix).