pR2 function

compute various pseudo-R2 measures

compute various pseudo-R2 measures for various GLMs

pR2(object, ...)

Arguments

• object: a fitted model object for which logLik, update, and model.frame methods exist (e.g., an object of class glm, polr, or multinom)
• ...: additional arguments to be passed to or from functions

Details

Numerous pseudo r-squared measures have been proposed for generalized linear models, involving a comparison of the log-likelihood for the fitted model against the log-likelihood of a null/restricted model with no predictors, normalized to run from zero to one as the fitted model provides a better fit to the data (providing a rough analogue to the computation of r-squared in a linear regression).

Returns

A vector of length 6 containing - llh: The log-likelihood from the fitted model

• llhNull: The log-likelihood from the intercept-only restricted model

• G2: Minus two times the difference in the log-likelihoods

• McFadden: McFadden's pseudo r-squared

• r2ML: Maximum likelihood pseudo r-squared

• r2CU: Cragg and Uhler's pseudo r-squared

References

Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables. Sage. pp104-106.

Author(s)

Simon Jackman simon.jackman@sydney.edu.au

See Also

extractAIC, logLik

Examples

data(admit)
## ordered probit model
op1 <- MASS::polr(score ~ gre.quant + gre.verbal + ap + pt + female,
            Hess=TRUE,
            data=admit,
            method="probit")
pR2(op1)