Dependences:
irr package, which was previously used to internally calculate Cohen's Kappa (irr::kappa2). Now, Cohen's Kappa is calculated using a native base implementation to improve stability and maintainability. In addition, the Cohen's Kappa calculation is now faster.Documentation:
roxygen2 (v8.0.0) standards.README.Rmd plot rendering.miclust() and its related methods print.miclust() and plot.miclust() into a single help page for a more fluid usage.summary.miclust() and print.summary.miclust().miclust:
print.miclust:
print.miclust.plot.miclust:
metric, to select which metric is displayed. Defaults to "all", maintaining compatibility with the output from versions 1.2.8 and earlier. Other options are: "nclfreq" (percentage of times each number of clusters has been selected); "critcf" (critCF distribution for each number of clusters); "nvarfreq" (distribution of the number of selected variables), and "varsel" (percentage of appearance of the variables that remained in the final set of selected variables).col.nclfreq, col.critcf, col.nvarfreq, and col.varsel, to customize the color used in the plots for metrics, "nclfreq", "critcf", "nvarfreq", and "varsel", respectively. Default values are consistent with results under versions 1.2.8 and earlier.col.all. When specified, it overrides col.nclfreq, col.critcf, col.nvarfreq, and col.varsel, and applies that color across the entire panel. Default (NULL) is consistent with results under versions 1.2.8 and earlier.summary.miclust:
summary.miclust() (originally triggered during kcca() execution inside assignprobandkappas()) that occurred when a specific imputation or the final assigned clustering result contained an empty cluster. Specifically, add a validation check in summary.miclust() to detect empty clusters across all imputations and the assigned cluster for the given k. The function now prints the affected cases and halts execution with a detailed error message.summary.miclust() (originally triggered during kcca() execution inside assignprobandkappas()) occurs when the final clustering for a given k and the set of selected variables results in an empty cluster within specific imputed datasets. This indicates that such a combination of k and selected variables is unable to find as many clusters as requested. A validation check has been added in summary.miclust() to detect these scenarios and halt execution with a detailed error message.print.summary.miclust:
Within-cluster summary section of the output has changed. To reduce the output width, the columns for missingness, mean, and standard deviation have been collapsed into a single column. The section label has been changed to Within-cluster summary [%miss.;mean;sd]. Separated columns (version <= 1.2.8 style) are still available in the summarybycluster element of the object.CITATION.README.