0
$\begingroup$

Assume we parametrize the variance in a negative binomial regression model as:

variance = mu(1+mu/theta).

Fitting the model with glmmTMB, I check theta from:

fixef(fit)$disp

In mgcv, I look at:

fit$family$getTheta()

Are those equivalent in meaning theta as in the variance formula?

$\endgroup$
1
  • $\begingroup$ I think I figured it out, though I can't seem to find official documentation on how to get the dispersion value in each and hope it corresponds to the variance parametrization. The dispersion in glmmTMB corresponds to the log of the dispersion in mgcv. $\endgroup$
    – Ilyco
    Apr 11, 2018 at 9:15

1 Answer 1

3
$\begingroup$

You can use the function sigma() to get the dispersion parameter (on the real rather than log scale). To interpret the dispersion parameters of any distribution, see ?sigma.glmmTMB.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.