README.md

In RobbyLankford/tidytest: Tidy Statistical Modeling Tests

tidytest

Overview

The goal of the tidytest package is to provide a tidy unified interface to common statistical tests, specifically those used when developing and validating statistical models.

Installation

You can install the development version of tidytest from GitHub with:

# install.packages("devtools")
devtools::install_github("RobbyLankford/tidytest")

Usage

The tidytest package is designed to be used with the tidymodels ecosystem, particularly the parsnip package. Currently, methods exist for the following parsnip models:

• linear_reg

library(parsnip)
library(tidytest)

mod_fit <- linear_reg() %>%
  set_engine("lm") %>%
  fit(mpg ~ disp + wt + hp, data = mtcars)

Tests in the tidytest package are designed to require minimal input. For most tests, the model object is the only required input.

Linear Tests

Tests for linear regression often focus on the behavior of the residuals. The classic assumptions for a linear model are the “LINE” assumptions of the residuals, specifically:

• Linearity
• Independence
• Normality
• Equal Variance

Linearity

Currently, the implemented linearity tests are:

• Ramsey RESET Test
• Harvey-Collier Test

ramsey_reset(mod_fit)
harvey_collier_test(mod_fit)

Independence

Currently, the implemented independence tests are:

• Durbin-Watson Test
• Ljung-Box Test

durbin_watson_test(mod_fit)
ljung_box_test(mod_fit)

Normality

Currently, the implemented normality tests are:

• Anderson-Darling Test
• Shapiro-Wilk Test

anderson_darling_test(mod_fit)
shapiro_wilk_test(mod_fit)

Equal Variance

Currently, the implemented equal variance (homoscedasticity) tests are:

• Bruesch-Pagan Test
• Goldfeld-Quandt Test

bruesch_pagan_test(mod_fit)
goldfeld_quandt_test(mod_fit)

RobbyLankford/tidytest documentation built on Jan. 27, 2024, 7:36 a.m.