Penguins Dataset

Description

A dataset that contains information from a study done on dives of emperor penguins by Jessica Meir and others. The scientists were interested in understanding how the heart rate of the penguins relates to the depth and duration of the dive. The study involved attaching a device to penguins that recorded the heart rate of the bird during a dive. The dataset contains multiple observations recorded from 9 penguins. The dataset has 125 observations and 4 variables.

Usage

penguins

Format

A data.frame.

Details

The variables in the dataset are as follows.

Source

https://dasl.datadescription.com/datafile/penguins/?_sfm_methods=Regression&_sfm_cases=4+59943&sf_paged=9

Panel of Diagnostic Residual Plots.

Description

Creates a panel of residual diagnostic plots given inputs of residuals and fitted values.

Usage

resid_auxpanel(residuals, predicted, plots = "default", bins = 30,
  smoother = FALSE, qqline = TRUE, qqbands = FALSE, scale = 1,
  theme = "bw", axis.text.size = 10, title.text.size = 12,
  title.opt = TRUE, nrow = NULL)

Arguments

Details

The following grid options can be chosen for the plots argument.

• "all": This creates a panel of all plot types included in the package that are available for resid_auxpanel. (See plot descriptions under individual options.)

• "default": This creates a panel with a residual plot, a normal quantile plot of the residuals, an index plot of the residuals, and a histogram of the residuals.

• "SAS": This creates a panel of a residual plot, a normal quantile plot of the residuals, a histogram of the residuals, and a boxplot of the residuals. This was modeled after the residpanel option in proc mixed from SAS version 9.4.

• A vector of individual plots can also be specified. For example, one can specify plots = c("boxplot", "hist") or plots = "qq". The individual plot options are as follows.

  • "boxplot": A boxplot of residuals

  • "hist": A histogram of residuals

  • "index": A plot of residuals versus observation number

  • "qq": A normal quantile plot of residuals

  • "resid": A plot of residuals versus predicted values

Details on the creation of the plots can be found in the details section of the help file for resid_panel.

Value

A panel of residual diagnostic plots containing plots specified.

Examples

# Fit a regression tree to the penguins data
penguin_tree <- rpart::rpart(heartrate ~ depth + duration, data = penguins)

# Obtain the predictions from the model on the observed data
penguin_tree_pred <- predict(penguin_tree)

# Obtain the residuals from the model
penguin_tree_resid <- penguins$heartrate - penguin_tree_pred

# Create a panel with the residual and index plot
resid_auxpanel(residuals = penguin_tree_resid,
             predicted = penguin_tree_pred,
             plots = c("resid", "index", "yvp"))

Panel of Diagnostic Residual Plots Across Multiple Models.

Description

Creates a panel of residual diagnostic plots given a list of models. Currently accepts models of type "lm", "glm", "lmerMod", "lmerModLmerTest", and "glmerMod".

Usage

resid_compare(models, plots = "default", type = NA, bins = 30,
  smoother = FALSE, qqline = TRUE, qqbands = FALSE, scale = 1,
  theme = "bw", axis.text.size = 10, title.text.size = 12,
  title.opt = TRUE, nrow = NULL)

Arguments

Details

The first two sections below contain information on the available input options for the plots and type arguments in resid_compare. The third section contains details relating to the creation of the plots.

Options for Plots

The following options can be chosen for the plots argument.

• "all": This creates a panel of all plot types included in the package that are available for the model type input into residpanel. (See note below.)

• "default": This creates a panel with a residual plot, a normal quantile plot of the residuals, an index plot of the residuals, and a histogram of the residuals.

• "R": This creates a panel with a residual plot, a normal quantile plot of the residuals, a location-scale plot, and a leverage versus residuals plot. This was modeled after the plots shown in R if the plot() base function is applied to an lm model. This option can only be used with an lm or glm model.

• "SAS": This creates a panel with a residual plot, a normal quantile plot of the residuals, a histogram of the residuals, and a boxplot of the residuals. This was modeled after the residpanel option in proc mixed from SAS version 9.4.

• A vector of individual plots can also be specified. For example, one can specify plots = c("boxplot", "hist") or plots = "qq". The individual plot options are as follows.

  • "boxplot": A boxplot of residuals

  • "cookd": A plot of Cook's D values versus observation numbers

  • "hist": A histogram of residuals

  • "index": A plot of residuals versus observation numbers

  • "ls": A location scale plot of the residuals

  • "qq": A normal quantile plot of residuals

  • "lev": A plot of leverage values versus residuals

  • "resid": A plot of residuals versus predicted values

  • "yvp":: A plot of observed response values versus predicted values

Note: "cookd", "ls", and "lev" are only available for "lm" and "glm" models.

Options for Type

Several residual types are available to be requested based on the model type that is input into resid_panel. These currently are as follows.

• lm residual options

  • "pearson":The Pearson residuals

  • "response": The raw residuals (Default for "lm")

  • "standardized": The standardized raw residuals

• glm residual options

  • "pearson": The Pearson residuals

  • "deviance": The deviance residuals (Default for "glm")

  • "response": The raw residuals

  • "stand.deviance": The standardized deviance residuals

  • "stand.pearson": The standardized Pearson residuals

• lmer, lmerTest, and lme residual options

  • "pearson": The Pearson residuals (Default for "lmer", "lmerTest", and "lme")

  • "response": The raw residuals

• glmer residual options

  • "pearson": The Pearson residuals

  • "deviance": The deviance residuals (Default for "glmer")

  • "response": The raw residuals

Note: The plots of "ls" and "lev" only accept standardized residuals.

Details on the Creation of Plots

Boxplot (boxplot)

Boxplot of the residuals.

Cook's D (cookd)

The horizontal line represents a cut-off to identify highly influential points. The horizontal line is placed at 4/n where n is the number of data points used in the model.

Histogram (hist)

Plots a histogram of the residuals. The density curve overlaid has mean equal to zero and standard deviation equal to the standard deviation of the residuals.

Index Plot (index)

Plots the residuals on the y-axis and the observation number associated with the residual on the x-axis.

Leverage Plot (lev)

Plots the standardized residuals on the y-axis and the leverage values on the x-axis. A lowess curve is overlaid, and Cook's D contours are included for \alpha = 0.5 and \alpha = 1.

Location-Scale Plot (ls)

Plots the square root of the absolute value of the standardized residuals on the y-axis and the predicted values on the x-axis. The predicted values are plotted on the original scale for glm and glmer models. A lowess curve is overlaid.

QQ Plot (qq)

Makes use of the R package qqplotr for creating a normal quantile plot of the residuals.

Residual Plot (resid)

Plots the residuals on the y-axis and the predicted values on the x-axis. The predicted values are plotted on the original scale for glm and glmer models.

Response vs. Predicted (yvp)

Plots the response variable from the model on the y-axis and the predicted values on the x-axis. Both response variable and predicted values are plotted on the original scale for glm and glmer models.

Value

A panel of residual diagnostic plots containing plots specified for each model.

Examples

# Fit two models to the penguins data
penguin_model <- lme4::lmer(heartrate ~ depth + duration + (1|bird), data = penguins)
penguin_model_log2 <- lme4::lmer(log(heartrate) ~ depth + duration + I(duration^2) +
(1|bird), data = penguins)

# Compare the residuals from the model
resid_compare(list(penguin_model, penguin_model_log2))

# Adjust some options in the panel of plots
resid_compare(list(penguin_model, penguin_model_log2), plots = c("resid", "yvp"),
smoother = TRUE, theme = "grey")

Panel of Interactive Versions of Diagnostic Residual Plots.

Description

Creates a panel of interactive residual diagnostic plots given a model. Currently accepts models of type "lm", "glm", "lmerMod", "lmerModLmerTest", "lme", and "glmerMod".

Usage

resid_interact(model, plots = "default", type = NA, bins = 30,
  smoother = FALSE, qqline = TRUE, scale = 0.9, theme = "bw",
  axis.text.size = 10, title.text.size = 12, title.opt = TRUE,
  nrow = NULL)

Arguments

Details

Details on the creation of the plots can be found in the details section of the help file for resid_panel.

Value

A panel of interactive residual diagnostic plots containing plots specified.

Examples

# Fit a model to the penguin data
penguin_model <- lme4::lmer(heartrate ~ depth + duration + (1|bird), data = penguins)

# Create the default interactive panel
resid_interact(penguin_model)

# Select all plots to include in the panel and set the smoother option to TRUE
resid_interact(penguin_model, plots = "all", smoother = TRUE)

# Select only the residual plot and qq-plot to be included in the panel and
# set the number of rows to 2
resid_interact(penguin_model, plots = c("resid", "qq"), nrow = 2)

# Choose the SAS panel of plots, change the theme to classic, and remove the
# titles of the plots
resid_interact(penguin_model, plots = "SAS", theme = "classic", title.opt = FALSE)

Panel of Diagnostic Residual Plots.

Description

Creates a panel of residual diagnostic plots given a model. Currently accepts models of type "lm", "glm", "lmerMod", "lmerModLmerTest", "lme", and "glmerMod".

Usage

resid_panel(model, plots = "default", type = NA, bins = 30,
  smoother = FALSE, qqline = TRUE, qqbands = FALSE, scale = 1,
  theme = "bw", axis.text.size = 10, title.text.size = 12,
  title.opt = TRUE, nrow = NULL)

Arguments

Details

The first two sections below contain information on the available input options for the plots and type arguments in resid_panel. The third section contains details relating to the creation of the plots.

Options for Plots

The following options can be chosen for the plots argument.

• "all": This creates a panel of all plot types included in the package that are available for the model type input into resid_panel. (See note below.)

• "default": This creates a panel with a residual plot, a normal quantile plot of the residuals, an index plot of the residuals, and a histogram of the residuals.

• "R": This creates a panel with a residual plot, a normal quantile plot of the residuals, a location-scale plot, and a residuals versus leverage plot. This was modeled after the plots shown in R if the plot() base function is applied to an lm model. This option can only be used with an lm or glm model.

• "SAS": This creates a panel with a residual plot, a normal quantile plot of the residuals, a histogram of the residuals, and a boxplot of the residuals. This was modeled after the residualpanel option in proc mixed from SAS version 9.4.

• A vector of individual plots can also be specified. For example, one can specify plots = c("boxplot", "hist") or plots = "qq". The individual plot options are as follows.

  • "boxplot": A boxplot of residuals

  • "cookd": A plot of Cook's D values versus observation numbers

  • "hist": A histogram of residuals

  • "index": A plot of residuals versus observation numbers

  • "ls": A location scale plot of the residuals

  • "qq": A normal quantile plot of residuals

  • "lev": A plot of standardized residuals versus leverage values

  • "resid": A plot of residuals versus predicted values

  • "yvp":: A plot of observed response values versus predicted values

Note: "cookd", "ls", and "lev" are only available for "lm" and "glm" models.

Options for Type

Several residual types are available to be requested based on the model type that is input into resid_panel. These currently are as follows.

• lm residual options

  • "pearson":The Pearson residuals

  • "response": The raw residuals (Default for "lm")

  • "standardized": The standardized raw residuals

• glm residual options

  • "pearson": The Pearson residuals

  • "deviance": The deviance residuals (Default for "glm")

  • "response": The raw residuals

  • "stand.deviance": The standardized deviance residuals

  • "stand.pearson": The standardized Pearson residuals

• lmer, lmerTest, and lme residual options

  • "pearson": The Pearson residuals (Default for "lmer", "lmerTest", and "lme")

  • "response": The raw residuals

• glmer residual options

  • "pearson": The Pearson residuals

  • "deviance": The deviance residuals (Default for "glmer")

  • "response": The raw residuals

Note: The plots of "ls" and "lev" only accept standardized residuals.

Details on the Creation of Plots

Boxplot (boxplot)

Boxplot of the residuals.

Cook's D (cookd)

The horizontal line represents a cut-off to identify highly influential points. The horizontal line is placed at 4/n where n is the number of data points used in the model.

Histogram (hist)

Plots a histogram of the residuals. The density curve overlaid has mean equal to zero and standard deviation equal to the standard deviation of the residuals.

Index Plot (index)

Plots the residuals on the y-axis and the observation number associated with the residual on the x-axis.

Residual-Leverage Plot (lev)

Plots the standardized residuals on the y-axis and the leverage values on the x-axis with a loess curve is overlaid. Cook's D contour lines (which are a function of leverage and standardized residuals) are plotted as the red dashed lines for Cook's D values of 0.5 and 1.

Location-Scale Plot (ls)

Plots the square root of the absolute value of the standardized residuals on the y-axis and the predicted values on the x-axis. The predicted values are plotted on the original scale for glm and glmer models. A loess curve is overlaid.

QQ Plot (qq)

Makes use of the R package qqplotr for creating a normal quantile plot of the residuals.

Residual Plot (resid)

Plots the residuals on the y-axis and the predicted values on the x-axis. The predicted values are plotted on the original scale for glm and glmer models.

Response vs. Predicted (yvp)

Plots the response variable from the model on the y-axis and the predicted values on the x-axis. Both response variable and predicted values are plotted on the original scale for glm and glmer models.

Value

A panel of residual diagnostic plots containing plots specified.

Examples

# Fit a model to the penguin data
penguin_model <- lme4::lmer(heartrate ~ depth + duration + (1|bird), data = penguins)

# Create the default panel
resid_panel(penguin_model)

# Select all plots to include in the panel and set the smoother option to TRUE
resid_panel(penguin_model, plots = "all", smoother = TRUE)

# Select only the residual plot and qq-plot to be included in the panel,
# request confidence bands on the qq plot, and set the number of rows to 2
resid_panel(penguin_model, plots = c("resid", "qq"), qqbands = TRUE, nrow = 2)

# Choose the SAS panel of plots, change the theme to classic, and remove the
# titles of the plots
resid_panel(penguin_model, plots = "SAS", theme = "classic", title.opt = FALSE)

Panel of Plots of Residuals or Response Variable versus Predictor Variables.

Description

Creates a panel of plots of the residuals or response variable versus the predictor (x) variables in the model. Interactions between predictor variables are not included. Currently accepts models of type "lm", "glm", "lmerMod", "lmerModLmerTest", "lme", and "glmerMod".

Usage

resid_xpanel(model, yvar = "residual", type = NA, smoother = FALSE,
  scale = 1, theme = "bw", axis.text.size = 10,
  title.text.size = 12, title.opt = TRUE, nrow = NULL)

Arguments

Details

Note that for x variables that are factors, the levels shown on the x-axis will be in the order that the levels are ordered in the dataframe. This can be adjusted by reordering the levels of the factor before the model is fit.

Value

A panel of plots of the residuals or response variable versus the predictor variables.

Examples

# Fit a model to the penguin data
penguin_model <- lme4::lmer(heartrate ~ depth + duration + (1|bird), data = penguins)

# Create plots of the residuals versus the predictor variables
resid_xpanel(penguin_model, theme = "classic")

# Create plots of the response variable versus the predictor variables
resid_xpanel(penguin_model, yvar = "response", theme = "classic", smoother = TRUE)