[ADMB Users] gradient_structure problem with nbinom glmm.admb in R

dave fournier davef at otter-rsch.com
Sun Jun 12 10:42:01 PDT 2011

The easiest way is to modify the glmm.admb command.

In R do something like

 > sink("myglmmadmb")

 > glmm.admb

 > sink()

then you will have a file named myglmmadmb with the R script in it.

You can see that  the ad mb function is invoked vis sthe R system() 

Ben has abstracted the inteface a bit so that
the line

    cmdoptions = paste("-maxfn 500", if (impSamp == 0)

creates the string to include the CLO.  For a one off yu can just edit 
this to soemthing like

     cmdoptions = paste("-maxfn 500 -mno 60000 -ams 5000000 ", if 
(impSamp == 0)

(You may need some kind of -ams option for a big problem).

Nicer would be to modify the function itself

function (formula, data, family = "poisson", link, corStruct = "diag",
     impSamp = 0, easyFlag = TRUE, zeroInflation = FALSE, ZI_kluge = FALSE,
     imaxfn = 10, save.dir = NULL, verbose = FALSE, myoptions = NULL)

and add  myoptions to cmdoptions. Then you could just put the stuff you need
in your R code. (And try out the hybrid mcmc as well.)

More information about the Users mailing list