Densify: An R package to reduce empty cells in dataframes of typological linguistic data

Abstract

The R package densify provides a procedure to prune input data frames containing empty cells (or cells with values {?} or {NA}) to denser sub-matrices with fewer empty cells. The pruning process trades off a series of variably weighted concerns, including data retention, coding density (proportion of non-empty cells) and taxonomic diversity of rows (representing for example phylogenetic relations). Users can adapt the relative weights given to these concerns through various parameters so that the densification process best fits their needs. As such, the software is useful for several purposes, including the densification of sparse input matrices and the subsampling of large input matrices according to a procedure that is sensitive to taxonomic structure.