[R-sig-ME] glmmADMB truncated distributions

[Ben Bolker]

Ross Boylan <ross at ...> writes:

> 
> How are they truncated?  I expected there might be a way to set the 
> truncation threshhold, but I don't see it in the docs.
> Ross Boylan
> 

  They're truncated at zero.

  One of the problems with developing a flexible system is that everyone
comes up with new uses for it: the original intent of the truncated 
distributions was just to make hurdle models possible.

  It would be possible but not completely trivial to adapt glmmADMB
to handle other truncation points -- you would need a straightforward
way to compute the cumulative distribution function of the response up
to the truncation point. Probably the best thing for people who want
to extend glmmADMB in these various ways is to learn how to use the
full version of AD Model Builder ...

  Ben Bolker