cp_gR2 {gspcr}

Compute generalized R-squared

Description

Computes the Cox and Snell generalized R-squared.

Usage

cp_gR2(ll_n, ll_f, n)

Arguments

ll_n

numeric vector of length 1 (or an object of class 'logLik') storing the log-likelihood of the null (restricted) model

ll_f

numeric vector of length 1 (or an object of class 'logLik') storing the log-likelihood of the full model

n

numeric vector of length 1 storing the sample size of the data used to estimate the models

Details

The Cox and Snell generalized R-squared is equal to the R-squared when applied to multiple linear regression. The highest value for this measure is 1 - exp(ll_n)^(2/n), which is usually < 1. The null (restricted) model must be nested within the full model.

Value

numeric vector of length 1 storing the computed Cox and Snell generalized R-squared.

Author(s)

Edoardo Costantini, 2023

References

Allison, P. D. (2014, March). Measures of fit for logistic regression. In Proceedings of the SAS global forum 2014 conference (pp. 1-13). Cary, NC: SAS Institute Inc.

Examples

# Fit a null model
lm_n <- lm(mpg ~ 1, data = mtcars)

# Fit a full model
lm_f <- lm(mpg ~ cyl + disp, data = mtcars)

# Compute generalized R2
gr2 <- cp_gR2(
    ll_n = as.numeric(logLik(lm_n)),
    ll_f = as.numeric(logLik(lm_f)),
    n = nobs(lm_f)
)