DImodelsVis (version 1.0.4)

model_diagnostics_data: Data preparation of diagnostics plots for regression models with compositional predictors

Description

This function prepares the data-frame with necessary attributes for creating regression diagnostics plots for a model with compositional predictors where points are replaced by pie-glyphs making it easier to track various data points in the plots. The output data-frame can be passed to model_diagnostics_plot to create the visualisation.

Usage

model_diagnostics_data(model, prop = NULL)

Value

The original data used for fitting the model with the response and all model predictors along with the following additional columns

.hat

Diagonal of the hat matrix.

.sigma

Estimate of residual standard deviation when corresponding observation is dropped from model.

.cooksd

The cook's distance (cooks.distance()) for each observation.

.fitted

Fitted values of model.

.resid

The residuals for the observations.

.stdresid

The standardised (Pearson) residuals for the observations.

Obs

A unique identifier for each observation.

Label

The labels to be displayed besides the observations in the plot.

.qq

The quantile values for the standardised residuals generated using qqnorm().

weights

The weights for each observation in the model (useful in the context of weighted regression).

Arguments

model

A statistical regression model object fit using lm, glm, nlme functions, etc.

prop

A character vector giving names of the compositional predictors in the model. If this is not specified then plots prepared using the data would not contain pie-glyphs.

Examples

library(DImodels)

## Load data
data(sim1)

## Fit model
mod1 <- lm(response ~ 0 + (p1 + p2 + p3 + p4)^2, data = sim1)

## Get data for diagnostics plot
diagnostics_data <- model_diagnostics_data(mod1,
                                           prop = c("p1", "p2", "p3", "p4"))
print(head(diagnostics_data))

## The compositional predictors in the data are added as attributes to the data
attr(diagnostics_data, "prop")