| NEWS | R Documentation |
JSS reference updated.
Principal axis transformation argument added to mds().
Correlation between X and Y-hat added to Procrustes().
sim2diss() fixed for missing similarities.
bugfix in unfolding() for user-specified starting solution.
stress0() is now an object of class "smacof".
Bug in smacofConstraint() for constraint.ties fixed.
svm_mdsplot() updated.
Nutshell vignettes added.
Alienation coefficient added to Procrustes().
Procrustes() returns pairwise distances.
jackknife() renamed to jackmds().
Legend option for jakcknife added to plot.smacofJK().
In smacofConstraint() "linear" constraint changed to "unrestricted".
In unfolding() "matrix" changed to "unconditional".
omega default in unfolding() set to 1.
Guttman datasets added (Guttman1965, Guttman1991).
Second rectangle dataset added (area-shape).
svm_mdsplot() added for facet plotting based on SVM fit.
Argument "scale" in biplotmds() added.
Color shading added to plot.icexplore().
Fixed column coordinates version of vmu() implemented.
Vignette added.
Help files for indvalues and PVQ40 data updated.
method.dat = "rows" as default in permtest.smacofR().
Slight acceleration in unfolding when using transformations.
Pseudo-confidence ellipses added via confEllipse().
Inverse MDS removed because it was buggy. Will be added later on again.
Extra sanity check added for fixed coordinates in unfolding added.
Parallelization is added for rowconditional unfolding.
External unfolding added.
In case of NA's, proper weight matrix is created in unfolding.
Vector unfolding model added.
Circular unfolding for all unfolding versions implemented.
variable names in morsescales corrected.
adjust argument in plot.driftvec added.
sanity check in smacofConstraint for diagonal C.
stability coefficient for bootstrap added.
print.imds() for inverseMDS() added.
Gravity computation slightly modified.
Vaziri-Xu dataset added.
Facial expression data added.
Legend arguments for Procrustes plots added.
some fixes in the output structure of inverseMDS().
biplotmds() and plot.mdsbi() fixed for single covariate vectors.
Additional conversion options added to sim2diss().
verbose output in uniscale().
transform() exported.
plot.smacofboot() updated.
unfolding() updated with relaxed update that often reduces the iterations by a factor 2).
unfolding() updated with an overal rescaling to the coordinates, distances, and d-hats.
smacofRect() updated to do nonlinear unfolding added matrix and row conditional).
prefscal() is an alias for unfolding() and smacofRect()
plot.smacofR() updated to show row transformations by smacofR and show column labels.
mspline transformation in mds() and smacofSym() now includes an intercept.
smacofRect updated for nonlinear unfolding added (only matrix conditional).
plot.smacofR updated to show transformations by smacofR.
Procrustes help file updated.
Bug in SPP fixed.
Some help files fixed.
Registration for C function calls fixed.
Plot option for summed distances for smacofIndDiff models added.
Fast optimal scaling in smacofSphere including msplines.
Primal algorithm in smacofSphere accelerated.
Bug for smacofSP Shepard plots fixed.
cex argument for unfolding configuration plot activated.
Thigmophobe label option in plot.mdsbi added.
Multiple panels for plot.smacofID improved (thanks to Lew Harvey for the fix).
type = "ratio" for all models restored.
Shepard plots for IndDiff-models and unfolding fixed. For IndDiff, a Shepard plot for each individual is produced.
New scale argument added for scaling the arrow endpoints (coefficients) in plot.mdsbi.
Canadian newspaper dataset added (word co-occurrences).
stress0() function added to compute stress for a particular (starting) configuration.
Bug in smacofIndDiff with missing input dissimilarities fixed.
icExplore() function incl. print/plot S3 added for initial configuration exploration.
KIPT dataset added.
Crime dataset added.
smacofConstraint() updated to allow for regional restrictions (diagonal case).
type = "interval" as new default in mds().
confdist instead of confdiss as output value in mds().
Labels in plot.smacof() are now positioned using thigmophobe.labels().
Rectangle data corrected, help file updated.
Convex hull option added to plot.smacof.
Age and gender added as attributes to PVQ40 dataset.
New what argument in plot.smacofR for row/column/joint plots added.
bootmds() incl. plot and print S3 added for bootstrapping MDS solutions.
Help file for perception data updated.
Bug in smacofConstraint() removed: now returns proper matrix C when constraint = "diagonal".
plot.smacofJK() improved.
Bug fix in plot.smacofID().
Aspect ratio 1 added as default for the bubble plot option in plot.smacof().
Function dissWeight added to create weights as a function of the dissimilarities. Removed dissimilarities with weights 0 from Shepard plot. multistart.smacof added.
In mds, smacofSym, smacofConstraint, smacofIndDiff, and smacofSphere the order of computations in update changed. This update improves the speed of computations for large data when using weightmat.
Hue change in jackknife plot.
Some fixes in the permutation tests.
Permutation tests based on permutations of the original data matrix added.
biplot function for creating MDS biplots added.
Optimization in smacofConstraint improved (monotone regression in two directions).
Some plot cosmetics for smacofIndDiff and Procrustes calls.
Transformations in smacofIndDiff use C implementation (mspline can be done as well).
Drift vector model added.
Datasets added.
smacofConstraint allows now for data frame and matrices for external variables, sanity checks added.
Verbose output now says "stress (raw)".
init default argument is now "torgerson" (before it was NULL which used Torgerson scaling anyway).
Bug for optimal scaling on external variables in smacofConstraint removed.
SPP-plot for unfolding models fixed.
Some datasets added.
Stress-per-point (spp) is now returned as percentage stress contribution including the squared error matrix (resmat).
If there are missing input dissimilarities, the corresponding weight matrix is created automatically.
Sanity check for negative input dissimilarities.
Some plot cosmetrics.
Output value structure for dual algorithm in smacofSphere changed, circle added in configuration plot.
Function randomstress for computing (average) stress values based on random input dissimilarities.
Unidimensional scaling added (all permutations): uniscale.
Procrustes added: Procrustes including print and plot methods.
Function torgerson (for classical scaling) is now exported.
Bug in smacofConstraint for external missing values removed.
Bug in smacofIndDiff with initial starting configurations fixed. Stress is now normalized to Stress-1.
Several wrapper functions added: mds, unfolding, indscal, idioscal.
Recoding of transformations:
Ordinal transformations are coded in C.
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).