phiWt: Binomial dispersion parameter.

In ABS-dev/PF: Prevented fraction

Binomial dispersion parameter.

Description

MME estimate of dispersion parameter phi.

Usage

phiWt(fit, subset.factor = NULL, fit.only = TRUE, show.warns = FALSE)

Arguments

fit	A glm object.
subset.factor	Factor for estimating phi by subset.
fit.only	Return only the new fit? If FALSE, also returns the weights and phi estimates.
show.warns	Show warnings

Details

Estimates binomial dispersion parameter \phi by the method of moments. Refits the model, weighting the observations by 1/\phi. Uses quasibinomial family in glm().

Value

A list with the following elements.

fit	the new model fit, updated by the estimated weights
weights	vector of weights
phi	vector of phi estimates

Author(s)

PF-package

References

Wedderburn RWM, 1974. Quasi-likelihood functions, generalized linear models, and the Gauss-Newton method. Biometrika 61:439-447.

See Also

tauWt, RRor.

Examples

birdm.fit <- glm(cbind(y, n - y)~tx-1, binomial, birdm)
RRor(phiWt(birdm.fit))
#
# 95% t intervals on 4 df
#
# PF
#     PF     LL     UL
#  0.479 -0.537  0.823
#
#       mu.hat   LL    UL
# txcon  0.768 0.95 0.367
# txvac  0.400 0.78 0.111
#

ABS-dev/PF documentation built on April 26, 2024, 3:29 p.m.