Computes (randomized) quantile residuals for lm and glm objects.
For lm objects, standardized residuals are returned. For glm objects,
the method depends on the family: binomial, poisson, Negative Binomial,
Bell, Gamma, inverse Gaussian; other families use a default
standardized deviance residual. Code adapted from the statmod package.
Usage
qresiduals(glm.obj, dispersion = NULL)
Arguments
- glm.obj
an object of class lm or glm.
- dispersion
optional dispersion parameter; if NULL, it is estimated
from the model. Only used for the Gamma, inverse Gaussian and default methods.
Value
A numeric vector with the quantile residuals.
Details
Residuals for discrete distributions are randomized, so results
vary between calls unless a seed is set.
References
Dunn, P. K. and Smyth, G. K. (1996). Randomized quantile
residuals. Journal of Computational and Graphical Statistics, 5, 236-244.
Author
Peter K. Dunn and Gordon Smyth (original code in statmod).