The RegAssure package is designed to simplify and enhance the process of validating regression model assumptions in R. It provides a comprehensive set of tools for evaluating key assumptions such as linearity, homoscedasticity, independence, normality, and collinearity, contributing to the reliability of analytical results.
You can easily install RegAssure from GitHub using the devtools package:
# install.packages("devtools")
devtools::install_github("nrubiog/RegAssure")Here’s a basic example showcasing how RegAssure can be used to enhance linear regression analysis:
# Install the package
# devtools::install_github("nrubiog/RegAssure")
# Load the package
library(RegAssure)
# Create a regression model
lm_model <- lm(mpg ~ wt + hp, data = mtcars)
# Check assumptions
check_lm_assumptions(lm_model)
#>
#> The assumption tests have been completed and the results are available in a list. Enjoy it :)
#> Las pruebas de supuestos han sido completadas y los resultados están disponibles en una lista. Disfrútalo :)
#> $Linearity
#> [1] 1.075529e-16
#>
#> $Homoscedasticity
#>
#> studentized Breusch-Pagan test
#>
#> data: model
#> BP = 0.88072, df = 2, p-value = 0.6438
#>
#>
#> $Independence
#>
#> Durbin-Watson test
#>
#> data: model
#> DW = 1.3624, p-value = 0.04123
#> alternative hypothesis: true autocorrelation is not 0
#>
#>
#> $Normality
#>
#> Shapiro-Wilk normality test
#>
#> data: model$residuals
#> W = 0.92792, p-value = 0.03427
#>
#>
#> $Multicollinearity
#> wt hp
#> 1.766625 1.766625Here’s an additional example demonstrating the use of RegAssure with logistic regression:
# Load the package
library(RegAssure)
library(titanic)
# Load the dataframe
titanic <- titanic_train
# Create a binary logistic regression model
logit_model <- glm(Survived ~ Pclass + Sex, data = titanic, family = "binomial")
# Check assumptions for binary logistic regression
check_logit(logit_model, data = titanic, tipo_modelo = "binario", vars_numericas = "Pclass", y = "Survived")
#> logit_model has no missing values.
#>
#> Tests performed for binary/binomial model.
#> Warning in check_logit(logit_model, data = titanic, tipo_modelo = "binario", : Box-Tidwell Test cannot be done.
#> Warning in check_logit(logit_model, data = titanic, tipo_modelo = "binario", : Variance Inflation Factor Test cannot be done.
#>
#> The assumption tests have been completed and the results are available in a list. Enjoy it :)
#> Las pruebas de supuestos han sido completadas y los resultados están disponibles en una lista. Disfrútalo :)
#> $model_type
#> [1] "binary/binomial"
#>
#> $Confusion
#> Predicciones
#> Variable 0 1
#> 0 468 81
#> 1 109 233
#>
#> $ROC
#>
#> Call:
#> roc.default(response = new_data[[y]], predictor = pred_logit, smooth = TRUE, auc = TRUE, ci = TRUE, ret = TRUE)
#>
#> Data: pred_logit in 549 controls (new_data[[y]] 0) < 342 cases (new_data[[y]] 1).
#> Smoothing: binormal
#> Area under the curve: 0.8453
#> 95% CI: 0.8115-0.8729 (2000 stratified bootstrap replicates)Here’s an example of how to use the get_predict() function to compare real and predicted values from a model:
# Load the package
library(RegAssure)
# Create a regression model
lm_model <- lm(mpg ~ wt + hp, data = mtcars)
# Get predictions and compare with real values
predictions <- get_predict(lm_model, mtcars, mtcars$mpg, n = 3)
# Print the results
print(predictions[1:7,])
#> reales predichos Error
#> Mazda RX4 21.0 23.572 -2.572
#> Mazda RX4 Wag 21.0 22.583 -1.583
#> Datsun 710 22.8 25.276 -2.476
#> Hornet 4 Drive 21.4 21.265 0.135
#> Hornet Sportabout 18.7 18.327 0.373
#> Valiant 18.1 20.474 -2.374
#> Duster 360 14.3 15.599 -1.299By incorporating RegAssure into your workflow, you can streamline the process of assessing and addressing regression model assumptions, leading to more informed decision-making.