tidy_categorical {tidycat}

Expand broom::tidy() Outputs for Categorical Parameter Estimates

Description

Create additional columns in a tidy model output (such as broom::tidy.lm()) to allow for easier control when plotting categorical parameter estimates.

Usage

tidy_categorical(
  d = NULL,
  m = NULL,
  include_reference = TRUE,
  reference_label = "Baseline Category",
  non_reference_label = paste0("Non-", reference_label),
  exponentiate = FALSE,
  n_level = FALSE
)

Arguments

d

A data frame tibble::tibble() output from broom::tidy.lm(); with one row for each term in the regression, including column term

m

A model object, created using a function such as lm()

include_reference

Logical indicating to include additional rows in output for reference categories, obtained from dummy.coef(). Defaults to TRUE

reference_label

Character string. When used will create an additional column in output with labels to indicate if terms correspond to reference categories.

non_reference_label

Character string. When reference_label is used will be in output to indicate if terms not corresponding to reference categories.

exponentiate

Logical indicating whether or not the results in broom::tidy.lm() are exponentiated. Defaults to FALSE.

n_level

Logical indicating whether or not to include a column n_level for the number of observations per category. Defaults to FALSE.

Value

Expanded tibble::tibble() from the version passed to d including additional columns:

variable

The name of the variable that the regression term belongs to.

level

The level of the categorical variable that the regression term belongs to. Will be an the term name for numeric variables.

effect

The type of term (main or interaction)

reference

The type of term (reference or non-reference) with label passed from reference_label. If reference_label is set NULL will not be created.

n_level

The the number of observations per category. If n_level is set NULL (default) will not be created.

In addition, extra rows will be added, if include_reference is set to FALSE for the reference categories, obtained from dummy.coef()

Author(s)

Guy J. Abel

See Also

broom::tidy.lm()

Examples

# strip ordering in factors (currently ordered factor not supported)
library(dplyr)
library(broom)

m0 <- esoph %>%
  mutate_if(is.factor, ~factor(., ordered = FALSE)) %>%
  glm(cbind(ncases, ncontrols) ~ agegp + tobgp * alcgp, data = .,
        family = binomial())
# tidy
tidy(m0)

# add further columns to tidy output to help manage categorical variables
m0 %>%
 tidy() %>%
 tidy_categorical(m = m0, include_reference = FALSE)

# include reference categories and column to indicate the additional terms
m0 %>%
 tidy() %>%
 tidy_categorical(m = m0)

# coefficient plots
d0 <- m0 %>%
  tidy(conf.int = TRUE) %>%
  tidy_categorical(m = m0) %>%
  # drop the intercept term
  slice(-1)
d0

# typical coefficient plot
library(ggplot2)
library(tidyr)
ggplot(data = d0 %>% drop_na(),
       mapping = aes(x = term, y = estimate,
                     ymin = conf.low, ymax = conf.high)) +
  coord_flip() +
  geom_hline(yintercept = 0, linetype = "dashed") +
  geom_pointrange()

# enhanced coefficient plot using additional columns from tidy_categorical and ggforce::facet_row()
library(ggforce)
ggplot(data = d0,
       mapping = aes(x = level, colour = reference,
                     y = estimate, ymin = conf.low, ymax = conf.high)) +
  facet_row(facets = vars(variable), scales = "free_x", space = "free") +
  geom_hline(yintercept = 0, linetype = "dashed") +
  geom_pointrange() +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))