[R-sig-ME] question on nbinom1

[Ben Bolker]

   I believe the R d/p/q/r functions corresponding to glmmTMB's 
implementation of *nbinom1 would look like this:


rnbinom1 <- function(n, mu, phi) {
     ## var = mu*(1+phi) = mu*(1+mu/k) -> k = mu/phi
     rnbinom(n, mu=mu, size=mu/phi)
}

dnbinom1 <- function(x, mu, phi, ...) {
     dnbinom(n, mu=mu, size=mu/phi, ...)
}

pnbinom1 <- function(q, mu, phi, ...) {
     pnbinom(q, mu=mu, size=mu/phi, ...)
}

qnbinom1 <- function(p, mu, phi, log=FALSE) {
     pnbinom(p, mu=mu, size=mu/phi, ...)
}


   (there would be an even more clever/inscrutable way to do this by 
transforming the body of the code, without repeating oneself so much, 
but it would probably be a bad idea)

On 10/12/20 6:34 AM, Mollie Brooks wrote:
> I think the easiest way to get a numerical representation of the distribution from a fitted model would be using the simulate function.
> 
> There’s an example of how to do that on pages 392-393 of this pdf (including Figs 6 and 7)
> https://journal.r-project.org/archive/2017/RJ-2017-066/RJ-2017-066.pdf <https://journal.r-project.org/archive/2017/RJ-2017-066/RJ-2017-066.pdf>
> 
> cheers,
> Mollie
> 
>> On 10Oct 2020, at 14:23, Don Cohen <don-lme4 using isis.cs3-inc.com> wrote:
>>
>> Mollie Brooks writes:
>>
>>> I don't have a copy of Hardin & Hilbe 2007 on hand, but I answered
>>> a few of your questions below.
>>
>> Thank you.
>>
>> One more question:
>>
>> How can I compute the nbinom1 distribution?
>> Is there a formula for the pdf or cdf ?  An R function ?
>>
> 
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-mixed-models using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>