GitHub - alforj/muir: Explore Data with Tree Data Structures (original) (raw)

#muir

The muir package allows users to explore a data.frame using the tree data structure with minimal effort by simply providing the data.frame columns to be explored. In addition, the muir package allows for more targeted tree data structures to be created with specific column criteria as a method for documenting and communicating the data structure and relationships within a data set. These data tree structures can be viewed within the RStudio console, standard browers, or saved as HTML for sharing using the htmlwidgets package.

The package leverages the infrastructure provided by DiagrammeR.

Installation

Install the development version of muir from GitHub using the devtools package.

devtools::install_github('alforj/muir')

Or, get the v0.1.0 release from CRAN.

Using muir

Basic example

A basic example using muir to explore the mtcars data set showing the 'cyl' and 'carb' columns and the relationship between them (in that order).

Providing the "*" qualifier (after the ":" separator) indicates that all distinct values in the column should be returned as separate nodes. Obviously, depending on the data this may result in a LOT of nodes. As a guard against this, especially during initial exploration, a node.limit parameter limits the number of nodes returned at each level and is defaulted to three (3). If there are more than 3 distinct values, then only the Top 3 occuring values will be returned. This global parameter can be adjusted for all node levels or specific limits can be provided individually for each node.level value (for example, by providing "cyl:4" instead of "cyl:*" to indicate a node limit of 4 for the 'cyl' column only).

The resulting tree will be rendered starting with a level 0 node counting all rows in the data set. Each resulting level will be based on the columns provided by the user in node.levels. Each subsequent level will carry the filters from previous parent nodes forward. Percentages will be provided for each node (compared to the level 0 count) by default and can be turned off if not desired.

tree.height and tree.width values control how the tree is rendered and can be adjusted to best fit trees of various depths and widths.

library(muir) data(mtcars) mtTree <- muir(data = mtcars, node.levels = c("cyl:", "carb:"), tree.height = 1200, tree.width = 800)

mtTree

More complicated example

Instead of (or in combination with) just returning top counts for columns provided in node.levels, you can provide custom filter criteria and custom node titles in level.criteria(level.criteria could also be read in from a stored file (e.g., a crtieria.csv) as a data.frame)

The criteria data.frame below includes the column names, operators, and associated values (e.g., "cyl" <= 4), and a node title to accompany each node generated for that filter criteria.

Adding a "+" suffix after the column name in the node.levels parameter will add an extra "Other" node that will aggregate all values not already provided in the level.criteria value or for all distinct values below the node.limit provided.

Additional label values can be provided with the label.vals parameter usingdplyr summary functions. Custom labels for each value can be provided by adding custom text after the ":" separator.

Lastly, the direction the tree is drawn can be changed from the default left-to-right to a top-to-bottom ("TB") rendering by providing a new value for tree.dir.

criteria <- data.frame(col = c("cyl", "cyl", "carb"), oper = c("<=", ">", "=="), val = c(4, 4, 2), title = c("Up to 4 Cylinders", "More than 4 Cylinders", "2 Carburetors"))

mtTree <- muir(data = mtcars, node.levels = c("cyl", "carb:+"), level.criteria = criteria, label.vals = c("n():Count", "min(wt):Min Weight", "max(wt):Max Weight"), tree.dir = "TB", tree.height = 400, tree.width = 800)

mtTree

More examples

More examples can be found in the examples in the muir function