Version 1.12-1: - Fixed bug in optimal scaling of disparities. Version 1.11-4: - Added a biplotmds method for pcops. Version 1.11-3: - Updated docs to include better hyperlinks. Also changed some solver names to match the function names we use. Touchups for CRAN release. Version 1.11-2: - Touchups for CRAN release. Version 1.11-1: - Fixed bug (or perhaps it intended at some point) where the returned OC is for the unscaled configuration with the dmax for the scaled config. We now return the scaled configuration as $conf, the unscaled as $usconf. OC now matches between manual calculation with foo$conf and summary function. Version 1.10-1: - Added jackmds and bootmds methods for pcops objects. Fixed calls in cop_* functions to work with jackmds and bootmds. Version 1.9-1: - Updates for CRAN. Fixed roxygen @docType change from here https://github.com/r-lib/roxygen2/issues/1491. Version 1.8-2: - Finished moving to smacofx functions as workhorse for all cop_X functions (mostly using scmaofx::powerStressMin). Removed rStressMin and powerStressMin. Version 1.8-0: - Forked all the functionality that was not directly cops related to smacofx. Now the main function is copStressMin which is essentially the cops version of a mash-up of smacofx::rstressMin and smacofx::powerStressMin, thus generalizing all the models in smacofx but lmds, bcmds and clca. cops-c models are of class smacofP and work with the infrastructure of smacofx. Similarly, objects returned from pcops have a slot fit that is of class smacofP or smacofB and thus can also use the smacofx functionality. Added type argument to all functions. Version 1.7-2: - Changed the Shepard plot to now work with power exponents in the dhat regression line added to the plot. The latter are the same one as obtained by the MDS fit only linearly transformed to be on the same scale as the configuration distances. Version 1.7-1: - Added bootmds, biplotmds, icExploreGen, multistart, permtest, jackmds function that work with smacofP and copsc objects. Version 1.7-0: - Completely redid the object class smacofP returned from the functions so that they match smacof objects (incl. all symmetric matrices are dist objects). New methods for generics of smacofP. updated functions to also allow for principal argument. Programmed some sanity checks. Weightmat can now also be a dist object. NA or +-inf in weightmat will be treated as 0. Version 1.6-1: - Changed Sammon mapping and elastic scaling functions optimal scaling to interval and ratio only. updated multiscale, alscal, rstressMin and powerStressMin. Moved many functions in their own files. Version 1.6-0: - Added Sammon mapping and elastic scaling functions that allows optimal scaling. Version 1.5-0: - Added a dedicated r-stress function rStressMin that supports ratio, interval and ordinal optimal scaling with powers for distances and weights. Power stress is now meant only for metric scaling with ratio and interval transformations and powers. Extended copStressMin to allow ordinal and interval optimal scaling be used with powers for distances and weights. Added multiscale and alscal functions. Version 1.4-3: - Added distance function for "Blended Chi-Square Distance" after Lindsay (1994). Exported the distance functions. Version 1.4-2: - Added aliasing functions copStressMin and copsc. Version 1.4-1: - Added a function for calculating the phi distance. Added aliasing functions copStressMin and copsc. Version 1.4-0: - Two-step algorithms now always use at least 0.1*itmax number of iterations for the second step even if itmax was exceeded by the first step, so itmax effectively is 1.1*itmax if there's no convergence prior. Fixed a bug when the number of iterations in the first step of a two-step algorithm exceeded itmax and a negative itmax was passed to the second algorithm. Now suppressing the useless warning from minqa:newuoa about maxfun < 10*length(par)^2 not recommended. Added (currently ineffective) code for using nloptr as engine for newuoa (instead of minqa) but not sure yet if we should switch (in test there was no clear advantage in terms of speed, accuracy and number of iterations). Fixed documentation and added details on the common error when two points are put on top of each other. Depend R version on 3.5.0 or higher due to matchphi data set. Version 1.3-5: - Added matchphi data. Changed default of itmax in powerStressMin, powerStressFast and copstressMin to 10000. Changed default of acc in powerStressFast and powerStressMin to 1e-6. Version 1.3-4: - Changed citations. Version 1.3-3: - Changed default of acc in powerStressMin to 1e-6. In powerStressFast to 1e-10. Version 1.3-2: - Removed torgerson function and use smacof::torgerson throughout. Version 1.3-1: - Added test to .Rbuildignore for the time being (test are 8 years out of date). Removed testthat from Suggests. Removed stats from Enhances. Changed docs. Added class of cmdscale wrapper to cmdscaleE and of sammon wrapper to sammonE and also cmdscale and sammon (to distinguish methods that are of general interest (cmdscale/sammon) or only within stops (cmdscaleE/sammonE). Tweaked Description. CRAN release. Version 1.3-0: - New version 1.3-0. Changed the theta arguments to only allow for the number of free parameters for P-COPS (as in STOPS). P-COPS models may also fitted with the stops package if one wants more/better solvers. Tweaks to docs. Version 1.2-1: - Updated some documentation for restricted power stress. Version 1.2-0: - Removed rgl functions. Updated reference to the JCGS publication. Fixed a bug in plot.cmdscale and plot.smacofP for plot.type="Shepard" and plot.type="resplot". Version 1.1-3: - Touchups for URLs and citation for CRAN. Version 1.1-2: - Set all standardisations for P-COPS functions to "sd". Version 1.1-1: - Added a vignette. Set all standardisations for P-COPS functions to "sd". Version 1.1-0: - Added rpowerstress, a powerstress model that restricts the transformations for dissimilarities and distances to be equal. Changed an (extra and not used) slot name of the pcops models from $cordillera to $copsobj. Updated docs. Version 1.0-0: - Updated docs, examples and description to pass CRAN checks. This is the first release. Version 0.8-0: - Changed optimizer default of copstressMin() to hjk-Newuoa and 5000 iterations as maximum. Version 0.7-0: - Added nonmetric COPS-C and docs. - Changed the summary and class of copsc and the according methods. - Added $typo to output of copstressMin Version 0.6-3: - Added ORCID. - Changed internals of all cops function to have a $theta and $parameters argument. copstress now extracts $parameters. print function for pcops models has been updated accordingly. Also, powerstressMin now returns $parameters, $theta as well. Version 0.6-2 - Changes to copstressMin: Stress normalization that is reported and used for copstress changed. Now the default reported and used is stress-1. Setting stresstype="stress" will use stress but also report stress-1. Version 0.6-1 - Changes to pcops: New initial configuration; now matches how powerstressMin does it. - Fixed bug in copsc that removed the rownames of the configuration and the delta of the object. Version 0.6-0 - Changes to all functions: Changed number of iteration defaults to much lower values (thanks to Patrick Groenen). At least a speedup of 2 is expected for COPS-C and about 10x for P-COPS. Version 0.5-0 - Changes to pcops: Added DIRECT and NOMAD optimization to pcops. Returned object is now called confdist instead of confdiss. Version 0.4-1 - Changes to pcops: Standardized the returned stresses. Reported is sqrt(stress) (object$stress) throughout but used for optimization is stress (object$stress.m). Version 0.4-0 - Changes to pcops: Added the approximated power stress which can be estimated with smacof and is much faster than powerstressMin. Version 0.3-0 - Changes to copstressMin: Added new deriv free optimizers. Version 0.2-0 - Changes to copstressMin: Added interval MDS for copstressMin. Version 0.1-3 - Changes to copstressMin: Changed the verbose levels. Added the dmax argument and handling (to be consistent with cordillera). Some tweaks to how to find the rang, dmax and init configuration (dmax and init if not supplied do now fit a torgerson model only) - much faster now as we no longer fit three powerstress MDS. Also, the configuration return under $conf now has the scaling from the scale argument, the unscaled configuration is now in the slot $confo. Stress is calculated for the explictly normalized config ($confo) but OC for the scaled config $conf. Scale="none" makes $conf and $confo equivalent. Changed the automatic dmax (now 1.5 times max reachability of power torgerson solution only if kappa>1, else 1 times max reachability). Version 0.1-2 - Added scaling of configs in pcops for OC calculation allows now "none", "sd" (divided by max standard deviation), "rmsq" (divided by maximum root mean square), "proc" (Procrustes adjustment to initial fit) and "std" (which is the old way and now discouraged). For this a new function is exported. Returned optimal configuration does not get scaled. - Adapted doc and removed the rstress example. Version 0.1-1 - Added new scaling of configs in copstressMin for OC calculation. Using scale() distorted the optimal config. Now one can choose between none, sd (divided by max standard deviation), rmsq (divided by maximum root mean square), proc (Procrustes adjustement to initial fit) and std (which is the old way and now discouraged). Added switch in copstressMin for choosing stress-1 and explicitly normed stress, with stress-1 now the default. Changed default stressweight and cordweight. Extended pcops to allow for the different scalings of X, changed default to the same as for cops-c. Now one can choose between TRUE or 1 which does centering/sd scaling the old way (each column differently now discouraged), 2 (scaling each column differently but no centering also not recommended when dealing with real distances), 3 (divided by max standard deviation), 4 (divided by maximum root mean square), FALSE or 0 for no scaling. Version 0.1-0 - Removed old versions and development versions of functions. - This is the package to be submitted with the paper. Incremented to 0.1-0. - Removed optics functions. Version 0.0-3 - Renamed all coploss to copstress. - Added a for loop version of powerstress (intended for optimization purposes eventually). - Removed all optics code. - Updated docs and descriptions. - Added some S3 methods for class pcops that were only in stops. Version 0.0-2 - Changed defaults for coplossMin to stressweight=0.99, cordweight=0.01. - Removed stops imports. - Removed cordillera fucntions. Version 0.0-1 - Forked COPS functionality to its own package