gam.nbut: Generalized Additive Models using Negative Binomial errors with unknown theta
Description
This is an internal function of package mgcv. It is called by gam()
for data where the error distribution is known to follow the Negative Binomial, but the
value of theta is unknown. Based on glm.nb() from the MASS library, it calls
gam.fit() repeatedly while iteratively improving its estimate of
theta, starting from a poisson error model and uses the neg.binom
error family (based again on code from the MASS library).Value
- returns a
gam object. See gam for details.