In SciViews/modelit: Statistical Models for 'SciViews::R'

knitr::opts_chunk$set(collapse = TRUE, comment = "#>", fig.align = 'center',
  fig.width = 7, fig.height = 5, out.width = "100%")
library(modelit)
library(tabularise)
library(chart)
library(dplyr)

The {modelit} package enhances statistical models calculated using lm(), glm() or nls() with nice charts and rich-formatted tables. It used both the fun$type(data = ...., formula) syntax. It also considers the labels of the variables in its outputs if a label attribute is defined.

Example

Load the required packages for this example.

library(modelit)
library(tabularise)
library(chart)
library(dplyr)

The Loblolly dataset is used to illustrate some {modelit} features. It focuses on the growth of Loblolly pine trees (Pinus tadea L. 1753). The dataset has three variables: height, age, and the seed.

data("Loblolly", package = "datasets")
# convert feet to meters
Loblolly$height <- round(Loblolly$height * 0.3048, 2)
# Add labels and units
Loblolly <- data.io::labelise(Loblolly,
  label = list(height = "Heigth", age = "Age", Seed = "Seed"),
  units = list(height = "m", age = "years"))

# First and last rows of the dataset with tabularise
tabularise$headtail(Loblolly)

Here is a plot of the data with chart(), automatically using labels and units defined here above:

chart(Loblolly, height ~ age %col=% Seed) +
  geom_line()

We only keep one measurement per tree (alias Seed) to get a dataset with independent observations to adjust a simple linear model.

set.seed(3652)
pine <- dplyr::slice_sample(Loblolly, n = 1, by = Seed)
chart(pine, height ~ age) + 
  geom_point()

We fit a quadratic model to the data.

pine_lm <- lm(data = pine, height ~ age + I(age^2))
summary(pine_lm)

Compare the "raw" summary of the lm object above with the version formatted using tabularise().

summary(pine_lm) |> 
  tabularise() 

The equation can be directly displayed in an R Markdown or Quarto document with the equation(pine_lm) code in an inline R chunk.

r equation(pine_lm)

Several plots are also available with chart(), starting with the graph showing the model fitted to the data:

chart(pine_lm) # Equivalent to : chart$model(pine_lm)

The plots of the analysis of the residuals are chart() variants that can be indicated in a short form as chart$<type>() function and the possible types are: "resfitted", "qqplot", "scalelocation", "cooksd", "resleverage", "cookleverage", "reshist", and "resautocor". A composite plot, incorporating the four most used residual analysis charts, is obtained with the "residuals" type (note that the four charts are labeled A-D to facilitate their description in the legend).

chart$residuals(pine_lm)

Nonlinear model

Other objects are also supported by {modelit}. For instance, a nonlinear model can be fitted to the same dataset. A Gompertz model is more suitable for such growth measurements.

pine_nls <- nls(data = pine, height ~ SSgompertz(age, a, b1, b2))
summary(pine_nls) |> 
  tabularise()

The plot of the model superposed to the data can be obtained once again with chart().

chart(pine_nls, lang = "fr")

SciViews/modelit documentation built on Nov. 24, 2024, 10:23 a.m.