Model Performance and Tests"

In JWileymisc: Miscellaneous Utilities and Functions

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

To start, load the package.

library(JWileymisc)

Model Performance

modelPerformance() is a generic function that can be used to calculate performance metrics for a model. JWileymisc implements methods for lm class objects. The output is a named list, with a data table containing results. For linear models, current performance metrics include:

• AIC (Akaike Information Criterion)
• BIC (Bayesian Information Criterion)
• LL (Log Likelihood)
• LLDF (degrees of freedom for log likelihood)
• Sigma (residual standard deviation)
• R2 ($R^2$ variance accounted for in the sample)
• F2 (Cohen's $f^2$ effect size, calculated as $\frac{R^{2}}{1 - R^{2}}$)
• AdjR2 (Sample size adjusted $R^2$, a better estimate of population variance accounted for)
• F (model F test)
• FNumDF (numerator degrees of freedom for model F test)
• FDenDF (denominator degrees of freedom for model F test)
• P (p value for model F test)

mtcars$cyl <- factor(mtcars$cyl)
m <- stats::lm(mpg ~ hp + cyl, data = mtcars)

mp <- modelPerformance(m)
print(mp)

If only certain metrics are desired, these can be found by extracting the "Performance" list element and then the correct column from the data table.

## Cohen's f^2 effect size
mp$Performance[, F2]

Another function, modelTest() is a generic providing a comprehensive series of tests for a model. Currently methods are implemented for both lm class models and vglm class models from the VGAM package with a multinomial family.

modelTest()

mt <- modelTest(m)
print(mt)

APAStyler(mt)

The model tests can also be used with interactions.

m2 <- stats::lm(mpg ~ hp * cyl, data = mtcars)

APAStyler(modelTest(m2))

JWileymisc documentation built on Oct. 5, 2023, 5:06 p.m.