In asshah4/forks: Formula Spells

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(nomnoml)
library(DiagrammeR)

Statement of Need

The raison d'étre of the {arcana} package ...

Development

The {arcana} package was developed as an extension of the formula class in base R. The {vctrs} package was used to extend the approach to formula handling.

Considerations

These are some of the thoughts that led to the development of this package.

The base formula class, as it applies to modeling and relationship descriptions
Expansion of formulas into groups/lists of related formulas
Fixed or protected roles of individual terms (e.g. outcomes, exposures, predictors, mediators, confounders)
Customized or unprotected roles for individual terms (specific for individual packages)
Operations that applies to individual terms and their distribution at the level of the data itself
Operations that apply to an overall formula itself, affecting the relationships between terms
Labels for individual terms (e.g. mtcars$cyl ~ "Number of Cylinders")

Vectors

As {vctrs} forms the base for this, a key concept is the idea of vector types and coercion hierarchy.

DiagrammeR::grViz("digraph {

graph [layout = dot, rankdir = BT]
node [shape = plaintext, fontname = arial]
edge [color = rosybrown]

# Terms
{
    rank=same
    ch [label = <<b>CHARACTER</b>>]
    ls [label = label]
    rl [label = role]
}
tx [label = <<b>TERM</b>>, fontcolor = crimson]
tl [label = <<b>TERM LIST</b>>, fontcolor = crimson]

# Formulas
fx [label = <<b>FORMULA</b>>, fontcolor = crimson]
pt [label = pattern]
fl [label = <<b>FORMULA LIST</b>>, fontcolor = crimson]

# Models
mx [label = <<b>MODEL</b>>]
ml [label = <<b>MODEL LIST</b>>, fontcolor = crimson]
df [label = 'data']

# Relationships
subgraph skeleton{
    ch -> tx
    tx -> ch 
    tx -> tl 
    tl -> tx 
    tl -> fx 
    fx -> tl 
    fx -> mx [weight = 0]
    fx -> fl [weight = 2]
    fl -> ml [weight = 2]
    mx -> ml [weight = 0]
}

ls -> tx [arrowhead = empty, weight = 0]
rl -> tx [arrowhead = empty, weight = 0]
pt -> fl [arrowhead = empty]
df -> mx [arrowhead = empty]
df -> ml [arrowhead = empty]

}")

In theory, the character class is considered enriched as a term class. In turn, a term class is considered enriched as a formula class. These enriched classes of term and formula could be implemented as vectors under this type of hierarchy, with emergent properties/attributes that develop in the hierarchy.

Structure

DiagrammeR::grViz("
digraph {

# Graph
graph [layout = dot, rankdir = RL]

# Nodes of object classes
term [shape = record, label='Term Record|+character|+side']

}")