Skip to contents

Ferrari's and Cribari-Neto's R2

Source: R/r2_ferarri.R
r2_ferrari.Rd

Calculates Ferrari's and Cribari-Neto's pseudo R2 (for beta-regression models).

Usage

r2_ferrari(model, ...)

# Default S3 method
r2_ferrari(model, correct_bounds = FALSE, ...)

Arguments

model

Generalized linear, in particular beta-regression model.

...

Currently not used.

correct_bounds

Logical, whether to correct the bounds of the response variable to avoid 0 and 1. If TRUE, the response variable is normalized and "compressed", i.e. zeros and ones are excluded.

Value

A list with the pseudo R2 value.

References

  • Ferrari, S., and Cribari-Neto, F. (2004). Beta Regression for Modelling Rates and Proportions. Journal of Applied Statistics, 31(7), 799–815. doi:10.1080/0266476042000214501

Examples

data("GasolineYield", package = "betareg")
model <- betareg::betareg(yield ~ batch + temp, data = GasolineYield)
#> Warning: NaNs produced
#> Error in while (testhalf & stepFactor < 11) {    fit <- fitfun(par, deriv = 2L)    scores <- gradfun(par, fit = fit)    InfoInv <- try(hessfun(par, fit = fit, inverse = TRUE))    if (failedInv <- inherits(InfoInv, "try-error")) {        warning("failed to invert the information matrix: iteration stopped prematurely")        break    }    bias <- if (type == "BR")         biasfun(par, fit = fit, vcov = InfoInv)$bias    else 0    par <- par + 2^(-stepFactor) * (step <- InfoInv %*% scores -         bias)    stepFactor <- stepFactor + 1    testhalf <- drop(crossprod(stepPrev) < crossprod(step))}: missing value where TRUE/FALSE needed
r2_ferrari(model)
#> # R2 for Generalized Linear Regression
#>   Ferrari's R2: 0.704