pixiedust (version 0.9.4)

tidy_levels_labels: Term and Level Descriptions for pixiedust Tables

Description

Default model objects identify rows of results with appropriate term name. More often than not, the term name is not suitable for formally reported output. tidy_levels_labels performs some basic work to quickly provide more readable descriptors for cases where they can easily be obtained. These descriptors are retrieved from the data, however, so the utility is determined by the user's habits in providing term labels and meaningful factor levels.

Due to the complexity of the terms that could be used for a model, it isn't practical to attempt to recover human-ready descriptors for every conceivable term. This would require recovering variable names for any number of functions. pixiedust only goes after the easiest to obtain. Replacements no managed by tidy_levels_labels may still be made with the replace sprinkle.

Usage

tidy_levels_labels(
  object,
  descriptors = "term",
  numeric_level = c("term", "term_plain", "label"),
  argcheck = NULL
)

Arguments

object

A model object, ideally with a model.frame method. It is unclear at the moment (18 Sept. 2015) what will happen if an object is passed that does not have a model.frame method.

descriptors

A character vector indicating the descriptors to be used in the table. Acceptable inputs are "term", "term_plain", "label", "level", and "level_detail". These may be used in any combination and any order, with the descriptors appearing in the table from left to right in the order given. The default, "term", returns only the term descriptor and is identical to the output provided by broom::tidy methods. See Details for a full explanation of each option and the Examples for sample output.

numeric_level

A character string that determines which descriptor is used for numeric variables in the "level_detail" descriptor when a numeric has an interaction with a factor. Acceptable inputs are "term", "term_plain", and "label".

argcheck

An assert collection created by checkmate::makeAssertCollection. Under normal circumstances, this is passed from dust. If NULL, as in the case it is run outside of dust, a new collection is created and the assertions are reported within tidy_levels_labels.

Restrictions

The descriptors, other than "term", generally don't make sense for data frame objects. The use of tidy_levels_labels is not permitted within the dust function, but is allowed if you really want it by pixiedust:::tidy_levels_labels.

Other special cases noted in future uses will be documented here, but in general, if it isn't a model object, you probably don't really want to use this.

Author

Benjamin Nutter

Details

The user may select up to five columns of descriptors, although doing so would certainly create some ambiguity. See the Examples for sample output.

• "term" The term name used in the R model summary

• "term_plain" The term name used in the formula. For variables that produce multiple term names (such as factors), the plain term name may be duplicated. For example, a factor that has term names FctrB and FctrC, indicating rows for levels B and C of the variable Fctr, will have two rows of "term_plain" of just Fctr.

• "label" Provides the label attached to the data using labelVector::get_label. When a term is not associated with a label, the value of term_plain is returned instead. Note that, variable names will disassociate with a label if they are used in a function (such as factor(x) or x^2.

• "level" Indicates the level being compared within a factor (or an interaction involving a factor), otherwise it returns NA. It may also be said that this value is the appendix to a factor name. For the term FctrB, this would just be B.

• "level_detail" Gives additional information to level by including the reference level of the factor. For the term FctrB, this would return "B vs A". When an interaction with a numeric variable is present, the level for the numeric may be either term_plain or label, the choice being controlled by the level_detail argument.

Examples

#* Descriptors for lm output with no interactions
mtcars2 <- mtcars
mtcars2$mpg <- labelVector::set_label(mtcars2$mpg, "Gas Mileage")
mtcars2$qsec <-  labelVector::set_label(mtcars2$qsec, "Quarter Mile Time")
mtcars2$am <-  labelVector::set_label(mtcars2$am, "Transmission")
mtcars2$wt <-  labelVector::set_label(mtcars2$wt, "Weight")
mtcars2$gear <-  labelVector::set_label(mtcars2$gear, "Gears")

#* Basic Output for a model with no interactions
#* Note: numeric_level has no impact as there are no
#*       interactions involving numeric variables.

fit <- lm(mpg ~ qsec + factor(am) + wt + factor(gear), data = mtcars2)

pixiedust:::tidy_levels_labels(fit, 
  descriptors = c("term", "term_plain", "label", "level", "level_detail"),
  numeric_level = "term") 
  
#* Assign factors ahead of the model. This allows 
#* the user to determine the levels that display.
#* Compare the output for 'am' with the output for 'gear'

mtcars2$am <- factor(mtcars2$am, 0:1, c("Automatic", "Manual"))
mtcars2$am <-  labelVector::set_label(mtcars2$am, "Transmission") 
    # Label was lost in variable conversion
fit <- lm(mpg ~ qsec + am + wt + factor(gear), data = mtcars2)
pixiedust:::tidy_levels_labels(fit, 
  descriptors = c("term", "term_plain", "label", "level", "level_detail"),
  numeric_level = "term") 
  
  
#* Include an interaction between a factor and numeric.

fit <- lm(mpg ~ qsec + am * wt + factor(gear), data = mtcars2)
pixiedust:::tidy_levels_labels(fit, 
  descriptors = c("term", "term_plain", "label", "level", "level_detail"),
  numeric_level = "term") 
  
#* Now observe how 'level' and 'level_detail' change 
#* in the interaction terms as we choose different 
#* values for 'numeric_level'

pixiedust:::tidy_levels_labels(fit, 
  descriptors = c("term", "term_plain", "label", "level", "level_detail"),
  numeric_level = "term_plain")
  
pixiedust:::tidy_levels_labels(fit, 
  descriptors = c("term", "term_plain", "label", "level", "level_detail"),
  numeric_level = "label")