coefplot: Plotting Model Coefficients

Description

Provides an S3 generic method for plotting coefficients from a model so it can be extended to other model types.

A graphical display of the coefficients and standard errors from a fitted model

coefplot is the S3 generic method for plotting the coefficients from a fitted model.

This can be extended with new methods for other types of models not currently available.

Coefplot method for workflow objects

Coefplot method for parsnip objects

Usage

coefplot(model, ...)
# S3 method for default
coefplot(
  model,
  title = "Coefficient Plot",
  xlab = "Value",
  ylab = "Coefficient",
  innerCI = 1,
  outerCI = 2,
  lwdInner = 1 + interactive * 2,
  lwdOuter = if (interactive) 1 else unname((Sys.info()["sysname"] != "Windows") * 0.5),
  pointSize = 3 + interactive * 5,
  color = "blue",
  shape = 16,
  cex = 0.8,
  textAngle = 0,
  numberAngle = 0,
  zeroColor = "grey",
  zeroLWD = 1,
  zeroType = 2,
  facet = FALSE,
  scales = "free",
  sort = c("natural", "magnitude", "alphabetical"),
  decreasing = FALSE,
  numeric = FALSE,
  fillColor = "grey",
  alpha = 1/2,
  horizontal = FALSE,
  factors = NULL,
  only = NULL,
  shorten = TRUE,
  intercept = TRUE,
  interceptName = "(Intercept)",
  coefficients = NULL,
  predictors = NULL,
  strict = FALSE,
  trans = identity,
  interactive = FALSE,
  newNames = NULL,
  plot = TRUE,
  ...
)
# S3 method for lm
coefplot(...)
# S3 method for glm
coefplot(...)
# S3 method for workflow
coefplot(model, ...)
# S3 method for model_fit
coefplot(model, ...)
# S3 method for rxGlm
coefplot(...)
# S3 method for rxLinMod
coefplot(...)
# S3 method for rxLogit
coefplot(...)

Arguments

model

A parsnip object

…

All arguments are passed on to coefplot.lm. Please see that function for argument information.

title

The name of the plot, if NULL then no name is given

xlab

The x label

ylab

The y label

innerCI

How wide the inner confidence interval should be, normally 1 standard deviation. If 0, then there will be no inner confidence interval.

outerCI

How wide the outer confidence interval should be, normally 2 standard deviations. If 0, then there will be no outer confidence interval.

lwdInner

The thickness of the inner confidence interval

lwdOuter

The thickness of the outer confidence interval

pointSize

Size of coefficient point

color

The color of the points and lines

shape

The shape of the points

cex

The text size multiplier, currently not used

textAngle

The angle for the coefficient labels, 0 is horizontal

numberAngle

The angle for the value labels, 0 is horizontal

zeroColor

The color of the line indicating 0

zeroLWD

The thickness of the 0 line

zeroType

The type of 0 line, 0 will mean no line

facet

logical; If the coefficients should be faceted by the variables, numeric coefficients (including the intercept) will be one facet. Currently not available.

scales

The way the axes should be treated in a faceted plot. Can be c("fixed", "free", "free_x", "free_y"). Currently not available.

sort

Determines the sort order of the coefficients. Possible values are c("natural", "magnitude", "alphabetical")

decreasing

logical; Whether the coefficients should be ascending or descending

numeric

logical; If true and factors has exactly one value, then it is displayed in a horizontal graph with continuous confidence bounds. Currently not available.

fillColor

The color of the confidence bounds for a numeric factor. Currently not available.

alpha

The transparency level of the numeric factor's confidence bound. Currently not available.

horizontal

logical; If the plot should be displayed horizontally. Currently not available.

factors

Vector of factor variables that will be the only ones shown

only

logical; If factors has a value this determines how interactions are treated. True means just that variable will be shown and not its interactions. False means interactions will be included.

shorten

logical or character; If FALSE then coefficients for factor levels will include their variable name. If TRUE coefficients for factor levels will be stripped of their variable names. If a character vector of variables only coefficients for factor levels associated with those variables will the variable names stripped. Currently not available.

intercept

logical; Whether the Intercept coefficient should be plotted

interceptName

Specifies name of intercept it case it is not the default of "(Intercept").

coefficients

A character vector specifying which factor coefficients to keep. It will keep all levels and any interactions, even if those are not listed.

predictors

A character vector specifying which coefficients to keep. Each individual coefficient can be specified. Use predictors to specify entire factors.

strict

If TRUE then predictors will only be matched to its own coefficients, not its interactions

trans

A transformation function to apply to the values and confidence intervals. identity by default. Use invlogit for binary regression.

interactive

If TRUE an interactive plot is generated instead of ggplot2

newNames

Named character vector of new names for coefficients

plot

logical; If the plot should be drawn, if false then a data.frame of the values will be returned

Value

A ggplot2 object or data.frame. See details in coefplot.lm for more information

If plot is TRUE then a ggplot object is returned. Otherwise a data.frame listing coefficients and confidence bands is returned.

A ggplot object. See coefplot.lm for more information.

Methods (by class)

default: Default method
lm: lm
glm: glm
workflow: tidymodels workflows
model_fit: parsnip
rxGlm: rxGlm
rxLinMod: rxLinMod
rxLogit: rxLogit

Details

Currently, methods are available for lm, glm and rxLinMod objects.

For more information on this function and it's arguments see coefplot.default

Pulls model element out of workflow object then calls coefplot.

Pulls model element out of parsnip object then calls coefplot.

Examples

Run this code

# NOT RUN {
data(diamonds)
head(diamonds)
model1 <- lm(price ~ carat + cut*color, data=diamonds)
model2 <- lm(price ~ carat*color, data=diamonds)
model3 <- glm(price > 10000 ~ carat*color, data=diamonds)
coefplot(model1)
coefplot(model2)
coefplot(model3)
coefplot(model1, predictors="color")
coefplot(model1, predictors="color", strict=TRUE)
coefplot(model1, coefficients=c("(Intercept)", "color.Q"))
coefplot(model1, predictors="cut", coefficients=c("(Intercept)", "color.Q"), strict=TRUE)
coefplot(model1, predictors="cut", coefficients=c("(Intercept)", "color.Q"), strict=FALSE)
coefplot(model1, predictors="cut", coefficients=c("(Intercept)", "color.Q"), 
strict=TRUE, newNames=c(color.Q="Color", "cut^4"="Fourth"))
coefplot(model1, predictors=c("(Intercept)", "carat"), newNames=c(carat="Size"))
coefplot(model1, predictors=c("(Intercept)", "carat"), 
newNames=c(carat="Size", "(Intercept)"="Constant"))


data(diamonds)
head(diamonds)
model1 <- lm(price ~ carat + cut*color, data=diamonds)
model2 <- lm(price ~ carat*color, data=diamonds)
coefplot(model1)
coefplot(model2)
coefplot(model1, predictors="color")
coefplot(model1, predictors="color", strict=TRUE)
coefplot(model1, coefficients=c("(Intercept)", "color.Q"))


model1 <- lm(price ~ carat + cut*color, data=diamonds)
coefplot(model1)

model2 <- glm(price > 10000 ~ carat + cut*color, data=diamonds, family=binomial(link="logit"))
coefplot(model2)
coefplot(model2, trans=invlogit)

# }
# NOT RUN {
mod4 <- rxGlm(price ~ carat + cut + x, data=diamonds)
mod5 <- rxGlm(price > 10000 ~ carat + cut + x, data=diamonds, family="binomial")
coefplot(mod4)
coefplot(mod5)
# }
# NOT RUN {

# }
# NOT RUN {
data(diamonds)
mod3 <- rxLinMod(price ~ carat + cut + x, data=diamonds)
coefplot(mod3)
# }
# NOT RUN {
# }
# NOT RUN {
data(diamonds)
mod6 <- rxLogit(price > 10000 ~ carat + cut + x, data=diamonds)
coefplot(mod6)
# }

Run the code above in your browser using DataLab