Plot model coefficients

ggcoef_model(), ggcoef_multinom() and ggcoef_compare() use broom.helpers::tidy_plus_plus() to obtain a tibble of the model coefficients, apply additional data transformation and then pass the produced tibble to ggcoef_plot() to generate the plot.

ggcoef_model(
  model,
  tidy_fun = broom.helpers::tidy_with_broom_or_parameters,
  tidy_args = NULL,
  conf.int = TRUE,
  conf.level = 0.95,
  exponentiate = FALSE,
  variable_labels = NULL,
  term_labels = NULL,
  interaction_sep = " * ",
  categorical_terms_pattern = "{level}",
  add_reference_rows = TRUE,
  no_reference_row = NULL,
  intercept = FALSE,
  include = dplyr::everything(),
  add_pairwise_contrasts = FALSE,
  pairwise_variables = broom.helpers::all_categorical(),
  keep_model_terms = FALSE,
  pairwise_reverse = TRUE,
  emmeans_args = list(),
  significance = 1 - conf.level,
  significance_labels = NULL,
  show_p_values = TRUE,
  signif_stars = TRUE,
  return_data = FALSE,
  ...
)

ggcoef_compare(
  models,
  type = c("dodged", "faceted"),
  tidy_fun = broom.helpers::tidy_with_broom_or_parameters,
  tidy_args = NULL,
  conf.int = TRUE,
  conf.level = 0.95,
  exponentiate = FALSE,
  variable_labels = NULL,
  term_labels = NULL,
  interaction_sep = " * ",
  categorical_terms_pattern = "{level}",
  add_reference_rows = TRUE,
  no_reference_row = NULL,
  intercept = FALSE,
  include = dplyr::everything(),
  add_pairwise_contrasts = FALSE,
  pairwise_variables = broom.helpers::all_categorical(),
  keep_model_terms = FALSE,
  pairwise_reverse = TRUE,
  emmeans_args = list(),
  significance = 1 - conf.level,
  significance_labels = NULL,
  return_data = FALSE,
  ...
)

ggcoef_multinom(
  model,
  type = c("dodged", "faceted"),
  y.level_label = NULL,
  tidy_fun = broom.helpers::tidy_with_broom_or_parameters,
  tidy_args = NULL,
  conf.int = TRUE,
  conf.level = 0.95,
  exponentiate = FALSE,
  variable_labels = NULL,
  term_labels = NULL,
  interaction_sep = " * ",
  categorical_terms_pattern = "{level}",
  add_reference_rows = TRUE,
  no_reference_row = NULL,
  intercept = FALSE,
  include = dplyr::everything(),
  significance = 1 - conf.level,
  significance_labels = NULL,
  show_p_values = TRUE,
  signif_stars = TRUE,
  return_data = FALSE,
  ...
)

ggcoef_plot(
  data,
  x = "estimate",
  y = "label",
  exponentiate = FALSE,
  point_size = 2,
  point_stroke = 2,
  point_fill = "white",
  colour = NULL,
  colour_guide = TRUE,
  colour_lab = "",
  colour_labels = ggplot2::waiver(),
  shape = "significance",
  shape_values = c(16, 21),
  shape_guide = TRUE,
  shape_lab = "",
  errorbar = TRUE,
  errorbar_height = 0.1,
  errorbar_coloured = FALSE,
  stripped_rows = TRUE,
  strips_odd = "#11111111",
  strips_even = "#00000000",
  vline = TRUE,
  vline_colour = "grey50",
  dodged = FALSE,
  dodged_width = 0.8,
  facet_row = "var_label",
  facet_col = NULL,
  facet_labeller = "label_value"
)

Arguments

model

a regression model object

tidy_fun

option to specify a custom tidier function

tidy_args

Additional arguments passed to broom.helpers::tidy_plus_plus() and to tidy_fun

conf.int

should confidence intervals be computed? (see broom::tidy())

conf.level

the confidence level to use for the confidence interval if conf.int = TRUE; must be strictly greater than 0 and less than 1; defaults to 0.95, which corresponds to a 95 percent confidence interval

exponentiate

if TRUE a logarithmic scale will be used for x-axis

variable_labels

a named list or a named vector of custom variable labels

term_labels

a named list or a named vector of custom term labels

interaction_sep

separator for interaction terms

categorical_terms_pattern

a glue pattern for labels of categorical terms with treatment or sum contrasts (see model_list_terms_levels())

add_reference_rows

should reference rows be added?

no_reference_row

variables (accepts tidyselect notation) for those no reference row should be added, when add_reference_rows = TRUE

intercept

should the intercept(s) be included?

include

variables to include. Accepts tidyselect syntax. Use - to remove a variable. Default is everything(). See also all_continuous(), all_categorical(), all_dichotomous() and all_interaction()

add_pairwise_contrasts

apply tidy_add_pairwise_contrasts()?

pairwise_variables

variables to add pairwise contrasts (accepts tidyselect notation)

keep_model_terms

keep original model terms for variables where pairwise contrasts are added? (default is FALSE)

pairwise_reverse

determines whether to use "pairwise" (if TRUE) or "revpairwise" (if FALSE), see emmeans::contrast()

emmeans_args

list of additional parameter to pass to emmeans::emmeans() when computing pairwise contrasts

significance

level (between 0 and 1) below which a coefficient is consider to be significantly different from 0 (or 1 if exponentiate = TRUE), NULL for not highlighting such coefficients

significance_labels

optional vector with custom labels for significance variable

show_p_values

if TRUE, add p-value to labels

signif_stars

if TRUE, add significant stars to labels

return_data

if TRUE, will return the data.frame used for plotting instead of the plot

...

parameters passed to ggcoef_plot()

models

named list of models

type

a dodged plot or a faceted plot?

y.level_label

an optional named vector for labeling y.level (see examples)

data

a data frame containing data to be plotted, typically the output of ggcoef_model(), ggcoef_compare() or ggcoef_multinom() with the option return_data = TRUE

x, y

variables mapped to x and y axis

point_size

size of the points

point_stroke

thickness of the points

point_fill

fill colour for the points

colour

optional variable name to be mapped to colour aesthetic

colour_guide

should colour guide be displayed in the legend?

colour_lab

label of the colour aesthetic in the legend

colour_labels

labels argument passed to ggplot2::scale_colour_discrete() and ggplot2::discrete_scale()

shape

optional variable name to be mapped to the shape aesthetic

shape_values

values of the different shapes to use in ggplot2::scale_shape_manual()

shape_guide

should shape guide be displayed in the legend?

shape_lab

label of the shape aesthetic in the legend

errorbar

should error bars be plotted?

errorbar_height

height of error bars

errorbar_coloured

should error bars be colored as the points?

stripped_rows

should stripped rows be displayed in the background?

strips_odd

color of the odd rows

strips_even

color of the even rows

vline

should a vertical line be drawn at 0 (or 1 if exponentiate = TRUE)?

vline_colour

colour of vertical line

dodged

should points be dodged (according to the colour aesthetic)?

dodged_width

width value for ggplot2::position_dodge()

facet_row

variable name to be used for row facets

facet_col

optional variable name to be used for column facets

facet_labeller

labeller function to be used for labeling facets; if labels are too long, you can use ggplot2::label_wrap_gen() (see examples), more information in the documentation of ggplot2::facet_grid()

Value

A ggplot2 plot or a tibble if return_data = TRUE.

Details

For more control, you can use the argument return_data = TRUE to get the produced tibble, apply any transformation of your own and then pass your customized tibble to ggcoef_plot().

Functions

• ggcoef_compare(): designed for displaying several models on the same plot.

• ggcoef_multinom(): a variation of ggcoef_model() adapted to multinomial logistic regressions performed with nnet::multinom().

• ggcoef_plot(): plot a tidy tibble of coefficients

Examples

mod <- lm(Sepal.Length ~ Sepal.Width + Species, data = iris)
ggcoef_model(mod)


# \donttest{
# a logistic regression example
d_titanic <- as.data.frame(Titanic)
d_titanic$Survived <- factor(d_titanic$Survived, c("No", "Yes"))
mod_titanic <- glm(
  Survived ~ Sex * Age + Class,
  weights = Freq,
  data = d_titanic,
  family = binomial
)

# use 'exponentiate = TRUE' to get the Odds Ratio
ggcoef_model(mod_titanic, exponentiate = TRUE)


# display intercepts
ggcoef_model(mod_titanic, exponentiate = TRUE, intercept = TRUE)


# customize terms labels
ggcoef_model(
  mod_titanic,
  exponentiate = TRUE,
  show_p_values = FALSE,
  signif_stars = FALSE,
  add_reference_rows = FALSE,
  categorical_terms_pattern = "{level} (ref: {reference_level})",
  interaction_sep = " x "
) +
  ggplot2::scale_y_discrete(labels = scales::label_wrap(15))
#> Scale for y is already present.
#> Adding another scale for y, which will replace the existing scale.


# display only a subset of terms
ggcoef_model(mod_titanic, exponentiate = TRUE, include = c("Age", "Class"))


# do not change points' shape based on significance
ggcoef_model(mod_titanic, exponentiate = TRUE, significance = NULL)


# a black and white version
ggcoef_model(
  mod_titanic, exponentiate = TRUE,
  colour = NULL, stripped_rows = FALSE
)


# show dichotomous terms on one row
ggcoef_model(
  mod_titanic,
  exponentiate = TRUE,
  no_reference_row = broom.helpers::all_dichotomous(),
  categorical_terms_pattern =
    "{ifelse(dichotomous, paste0(level, ' / ', reference_level), level)}",
  show_p_values = FALSE
)

# }

# \donttest{
data(tips, package = "reshape")
mod_simple <- lm(tip ~ day + time + total_bill, data = tips)
ggcoef_model(mod_simple)


# custom variable labels
# you can use the labelled package to define variable labels
# before computing model
if (requireNamespace("labelled")) {
  tips_labelled <- tips %>%
    labelled::set_variable_labels(
      day = "Day of the week",
      time = "Lunch or Dinner",
      total_bill = "Bill's total"
    )
  mod_labelled <- lm(tip ~ day + time + total_bill, data = tips_labelled)
  ggcoef_model(mod_labelled)
}


# you can provide custom variable labels with 'variable_labels'
ggcoef_model(
  mod_simple,
  variable_labels = c(
    day = "Week day",
    time = "Time (lunch or dinner ?)",
    total_bill = "Total of the bill"
  )
)

# if labels are too long, you can use 'facet_labeller' to wrap them
ggcoef_model(
  mod_simple,
  variable_labels = c(
    day = "Week day",
    time = "Time (lunch or dinner ?)",
    total_bill = "Total of the bill"
  ),
  facet_labeller = ggplot2::label_wrap_gen(10)
)


# do not display variable facets but add colour guide
ggcoef_model(mod_simple, facet_row = NULL, colour_guide = TRUE)


# works also with with polynomial terms
mod_poly <- lm(
  tip ~ poly(total_bill, 3) + day,
  data = tips,
)
ggcoef_model(mod_poly)


# or with different type of contrasts
# for sum contrasts, the value of the reference term is computed
if (requireNamespace("emmeans")) {
 mod2 <- lm(
   tip ~ day + time + sex,
   data = tips,
   contrasts = list(time = contr.sum, day = contr.treatment(4, base = 3))
  )
  ggcoef_model(mod2)
}
#> Loading required namespace: emmeans

# }

# \donttest{
# Use ggcoef_compare() for comparing several models on the same plot
mod1 <- lm(Fertility ~ ., data = swiss)
mod2 <- step(mod1, trace = 0)
mod3 <- lm(Fertility ~ Agriculture + Education * Catholic, data = swiss)
models <- list(
  "Full model" = mod1,
  "Simplified model" = mod2,
  "With interaction" = mod3
)

ggcoef_compare(models)

ggcoef_compare(models, type = "faceted")


# you can reverse the vertical position of the point by using a negative
# value for dodged_width (but it will produce some warnings)
ggcoef_compare(models, dodged_width = -.9)
#> Warning: `position_dodge()` requires non-overlapping x intervals
#> Warning: `position_dodge()` requires non-overlapping x intervals
#> Warning: `position_dodge()` requires non-overlapping x intervals
#> Warning: `position_dodge()` requires non-overlapping x intervals
#> Warning: `position_dodge()` requires non-overlapping x intervals
#> Warning: `position_dodge()` requires non-overlapping x intervals

# }

# \donttest{
# specific function for nnet::multinom models
mod <- nnet::multinom(Species ~ ., data = iris)
#> # weights:  18 (10 variable)
#> initial  value 164.791843 
#> iter  10 value 16.177348
#> iter  20 value 7.111438
#> iter  30 value 6.182999
#> iter  40 value 5.984028
#> iter  50 value 5.961278
#> iter  60 value 5.954900
#> iter  70 value 5.951851
#> iter  80 value 5.950343
#> iter  90 value 5.949904
#> iter 100 value 5.949867
#> final  value 5.949867 
#> stopped after 100 iterations
ggcoef_multinom(mod, exponentiate = TRUE)
#> ! `broom::tidy()` failed to tidy the model.
#> ✔ `tidy_parameters()` used instead.
#> ℹ Add `tidy_fun = broom.helpers::tidy_parameters` to quiet these messages.

ggcoef_multinom(mod, type = "faceted")
#> ! `broom::tidy()` failed to tidy the model.
#> ✔ `tidy_parameters()` used instead.
#> ℹ Add `tidy_fun = broom.helpers::tidy_parameters` to quiet these messages.

ggcoef_multinom(
  mod,
  type = "faceted",
  y.level_label = c("versicolor" = "versicolor\n(ref: setosa)")
)
#> ! `broom::tidy()` failed to tidy the model.
#> ✔ `tidy_parameters()` used instead.
#> ℹ Add `tidy_fun = broom.helpers::tidy_parameters` to quiet these messages.

# }